METHOD AND SYSTEM FOR WIRELESS ULTRA-LOW FOOTPRINT BODY SCANNING
1. A system for scanning and measuring a surface of a body of a person, comprising:
- a processor;
two or more range cameras capable of capturing at least a first depth image of a surface of a body from at least a first point of view while the body is in a pose, and at least a second depth image of the surface of the body from at least a second point of view while the body remains in the same pose, at least one of the two or more range cameras positioned at a different angle with respect to the other range cameras;
a first set of computer instructions executable on the processor capable of calculating a first set of three dimensional points from the first depth image and a second set of three dimensional points from the second depth image;
a second set of computer instructions executable on the processor capable of reconstructing the surface of the body by transforming the first and second set of three dimensional points into a final combined set of three dimensional points representing the surface of the body in a commonly shared coordinate system;
a third set of computer instructions executable on the processor capable of identifying points of measure from the final combined set of three dimensional points; and
a fourth set of computer instructions executable on the processor capable of extracting anthropometric body measurements using the points of measure.
A system for scanning a body to create scan data comprises a processor, a range camera capable of capturing at least a first set of depth images of the body rotated to 0 degrees, and at least a second set of depth images of the body rotated to x degrees, wherein x is >0 degrees, and x<360 degrees. a set of computer instructions are executable on a processor capable of synchronizing scan data and body measurements with a server.
- 1. A system for scanning and measuring a surface of a body of a person, comprising:
a processor; two or more range cameras capable of capturing at least a first depth image of a surface of a body from at least a first point of view while the body is in a pose, and at least a second depth image of the surface of the body from at least a second point of view while the body remains in the same pose, at least one of the two or more range cameras positioned at a different angle with respect to the other range cameras; a first set of computer instructions executable on the processor capable of calculating a first set of three dimensional points from the first depth image and a second set of three dimensional points from the second depth image; a second set of computer instructions executable on the processor capable of reconstructing the surface of the body by transforming the first and second set of three dimensional points into a final combined set of three dimensional points representing the surface of the body in a commonly shared coordinate system; a third set of computer instructions executable on the processor capable of identifying points of measure from the final combined set of three dimensional points; and a fourth set of computer instructions executable on the processor capable of extracting anthropometric body measurements using the points of measure.
- View Dependent Claims (2, 3, 4, 5, 6, 7, 8)
- 9. A method for scanning and measuring a surface of a body of a person, comprising:
capturing, using two or more range cameras, at least a first depth image of a surface of a body, from at least a first point of view while the body is in a pose, and at least a second depth image of the surface of the body, from at least a second point of view, while the body remains in the same pose, at least one of the two or more range cameras positioned at a different angle with respect to the other range cameras; calculating a first set of three dimensional points from the first depth image and a second set of three dimensional points from the second depth image; reconstructing the surface of the body by transforming the first and second set of three dimensional points into a final combined set of three dimensional points representing the surface of the body in a commonly shared coordinate system; identifying points of measure from the final combined set of three dimensional points; and extracting anthropometric body measurements from the points of measure.
- View Dependent Claims (10, 11, 12, 13, 14, 15, 16)
- 17. A system for scanning and measuring a surface of a body of a person, comprising:
a processor; a range camera capable of capturing a plurality of sets of depth images of a surface of a body, a first set of depth images comprising depth images of the surface of the body captured from at least a first point of view while the body is in a pose, each subsequent set of depth images comprising captured depth images of the surface of the body captured from at least a second point of view, while the body remains in the same pose, at least one of the two or more range cameras positioned at a different angle with respect to the other range cameras; a first set of computer instructions executable on the processor capable of calculating a set of three dimensional points for each set of depth images; a second set of computer instructions executable on the processor capable of reconstructing the surface of the body by transforming all sets of three dimensional points into one final set of three dimensional points representing the surface of the body in a commonly shared coordinate system; and a third set of computer instructions executable on the processor capable of identifying points of measure from the combined set of three dimensional points; a fourth set of computer instructions executable on the processor capable of extracting body measurements using the points of measure; and a wireless turntable having a rotatable platform capable of rotating the body to provide the two or more points of view of the body, wherein the processor as capable of wirelessly controlling rotation of the platform while the depth images are being captured; wherein the wireless turntable further comprises a scale capable of communicating with the processor to provide weight data for the body.
- 18. -32. (canceled)
This application is a Continuation of U.S. application Ser. No. 15/914,953, entitled “Method and System for Wireless Ultra-Low Footprint Body Scanning”, filed Mar. 7, 2018, which claims the benefit of Provisional Application Ser. No. 62/467,997, entitled “Method And System For Wireless Ultra-Low Footprint Body Scanning”, filed Mar. 7, 2017, and is a Continuation-In-Part of application of U.S. application Ser. No. 14/941,144, entitled “Cloud Server Body Scan Data System”, filed Nov. 13, 2015, which is a Continuation-In-Part of U.S. application Ser. No. 13/159,401, entitled “System And Method For Body Scanning And Avatar Creation”, filed Jun. 13, 2011, which is a Continuation-In-Part of U.S. application Ser. No. 13/008,906 filed Jan. 19, 2011 entitled “System And Method For 3d Virtual Try-On Of Apparel On An Avatar,” which claims the benefit of Application Ser. No. 61/352,390, entitled “System And Method For 3D Virtual Try-On Of Apparel On An Avatar”, filed Jun. 8, 2010, the contents of which are incorporated in this disclosure by reference in their entirety.
A portion of the disclosure of this patent document contains material that is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent document or the patent disclosure, as it appears in the Patent and Trademark Office patent file or records, but otherwise reserves all copyright rights whatsoever.
The invention relates to a method and system for wireless ultra-low footprint body scanning. More specifically, a scanning system enables a reduction in the footprint of a body scanner by using multiple stationary cameras on a single tower instead of a single camera that is stationary or moves on an adjoining rail.
Some of the major challenges to widespread adoption of body scanning in the clothing, fitness, health, medical, and other industries, has included the following, but is not limited to: (1) they use too large a footprint, (2) they are too difficult to use or operate, (3) they are too expensive, and (4) they lack portability. These limitations have made it hard for businesses of all sizes to adopt widely to body scanning. For example, a large footprint makes it more challenging for a commercial enterprise to invest in body scanning technology due to the high price of retail or commercial space. Previous body scanners that use multiple range cameras to scan a person have been limited to these large footprints. The large footprint is the result of placing many range cameras around a person. Improvements on these scanners have been made to reduce the footprint by introducing a turntable such that the person spins along 1-axis, while being captured by 1 range camera, such as the scanner from Styku, LLC of Los Angeles, Calif. Further improvements to Styku'"'"'s scanner had been to reduce the footprint further by placing a 1 range camera on a rail, effectively increasing its field of view, and thereby shortening the distance from the tower that encapsulates the rail and the turntable. Moreover, prior art body scanners with turntables would require some wired connection, therefore requiring more space to be occupied either by way of a wire, a cable management system, or a joining hardware unit between the turntable and camera tower, wherein a wired connection was made within the joining unit.
There is a need for scanner that requires a shorter footprint. The present invention solves that need.
In order to solve the problems and shortcomings of the prior art, an apparatus is disclosed for 3D virtual try-on of apparel on an avatar. According to one preferred embodiment, the system for 3D virtual try-on of apparel on an avatar is disclosed. According to one preferred embodiment, a method of fitting a garment on a person'"'"'s body online comprises receiving specifications of a garment, receiving body specifications of one or more fit models, receiving one or more grade rules, receiving one or more fabric specifications, and receiving specifications of a consumer'"'"'s body. The value of one or more fabric constants are determined according to the received one or more fabric specifications. One or more virtual garments in graded sizes are created and stored in a database based on the received garment specifications and fabric constants. Moreover, one or more graded virtual fit models are created and stored in a database based on the received specifications of the fit model. Each virtual garment is draped on the related virtual fit model to create a fit-model drape. An avatar is received or created to represent a consumer'"'"'s body shape. A selected one of the virtual garments is determined that represents a closest size for fitting on the avatar. The selected virtual garment is then re-draped on the consumer avatar. The consumer drape can then be viewed in 3D on the web or in a software application on any computing device. Data regarding the result of the virtual try-on process can then be utilized by the retailer, the consumer, and/or a third party. This virtual try-on data can be in the form of visual data or quantitative data that can be interpreted to determine the goodness of a garment'"'"'s fit. Specifically, consumers can be presented with such data to assess the appropriate size and the goodness of a garment'"'"'s fit, retailers can utilize such data for assessing how their garments are performing on their customer'"'"'s bodies, and finally, such data can be used as a predictive tool for recommending further garments to consumers (e.g., in a predictive, search or decision engine).
In another preferred embodiment, a method of fitting a garment on a person'"'"'s body online comprises receiving specifications of a garment, receiving specifications of a fit model, receiving a digital pattern corresponding to the fit model, receiving one or more grade rules, and receiving one or more fabric specifications. One or more graded digital patterns corresponding to one or more available sizes are calculated and stored in a database based on the received specifications of the garment, the received specifications of the fit model, the received digital pattern corresponding to the fit model, and the grade rules. The value of one or more fabric constants are determined according to the received one or more fabric specifications. An avatar representing the person'"'"'s body, and a selected one of the available sizes is determined that represents a closest size for fitting on the avatar. A virtual garment is created from the stored graded digital pattern corresponding to selected available size. The selected virtual garment is then draped on the avatar according to the fabric constants.
According to yet another preferred embodiment, a method of fitting a garment on a person'"'"'s body online comprises receiving specifications of a garment, receiving specifications of a fit model, receiving one or more grade rules, and receiving one or more fabric specifications. A virtual fit model is calculated and stored based on the received specifications of the garment, and the received specifications of the fit model. The values of one or more fabric constants are determined according to the received one or more fabric specifications. An avatar representing the person'"'"'s body is received, and a selected size for the person'"'"'s body is determined according to the received one or more grade rules. A virtual garment is created in the selected size according to the virtual fit model, the one or more grade rules, and the selected size. The selected virtual garment is then draped on the avatar according to the fabric constants.
In yet another preferred embodiment, a computer program product is stored on computer readable medium containing executable software instructions for fitting one or more garments on a person'"'"'s body, the executable software instructions.
In yet another preferred embodiment, a system for scanning a body comprises a processor, a range camera capable of capturing at least a first set of depth images of the body rotated to 0 degrees, and at least a second set of depth images of the body rotated to x degrees, wherein x is >0 degrees, and x<360 degrees, a first set of computer instructions executable on the processor capable of calculating a first set of three dimensional points from the first set of depth images and a second set of three dimensional points from the second set of depth images, a second set of computer instructions executable on the processor capable of rotating and translating the first and second set of three dimensional points into a final set of three dimensional points; and a third set of computer instructions executable on the processor capable of creating a three dimensional mesh from the final set of three dimensional points.
In yet another preferred embodiment, a scanning device provides yet another improvement on body scanning technology. The device described enables a further reduction in the footprint of a body scanner by using multiple stationary cameras on a single tower instead of a single camera that is stationary or moves on an adjoining rail. Using multiple stationary camera'"'"'s in combination of a turntable, reduces the distance between the camera tower, such that the overall footprint is the lowest ever created by a body scanning system. Keeping the cameras stationary reduces the complexity associated with moving parts on a rail. Moreover, the addition of a bluetooth or wireless network connected turntable that connects to an embedded PC within the range camera tower removes the need for a wired connection between the turntable and the range camera tower. The removal of all hard-wired connections between the turntable and the range camera tower further reduces the footprint, and improves convenience of installation, handling, and maintenance. Adding a voice assistant, LED lights, feature recognition, and a weight scale enable a plug and play experience that is interactive, intuitive, and enables the use of the scanner as a kiosk or home device.
For the purpose of illustrating the invention, there is shown in the accompanying drawings several embodiments of the invention. However, it should be understood by those of ordinary skill in the art that the invention is not limited to the precise arrangements and instrumentalities shown therein and described below.
The system for online virtual try-on of apparel on an avatar is disclosed in accordance with preferred embodiments of the present invention is illustrated in
A three dimensional (3D) virtual apparel processing system 112 gathers all or any combination of the following data available from retailer 50: (1) paper pattern 51, (2) grading rules 53, (3) technical pack 54, (4) digital pattern 57, (5) fit model'"'"'s scan data or measurements 58, (6) production sample garment, or (7) fabric swatches, where data displayed in
With reference to
In another embodiment, some retailers 50 may not have access to some of the information described above. For example, the retailer may not have any information on the pattern other than the technical pack 54, in which case a production sample garment 59 and technical pack 54 will be used by the 3D virtual apparel processing system 112. In another example, the retailer 50 may not provide a technical pack 54, in which case the production sample garment 59 is used for processing as described below.
In any case, whether a pattern, and/or technical pack 54 is received electronically from the producer'"'"'s digital garment data storage 52, or the less sophisticated garment information 60 is received, the information is processed into 3D virtual apparel processing system 112, and stored in a first data storage 110. In one embodiment, if the digital pattern 57 is received, it is imported into apparel product development software 114, and, if necessary, converted into the proper format. In another embodiment, if the patterns are not digital, they are digitized using a digitizer known to those skilled in the art. In another embodiment, if no pattern is received, then the pattern is made from the production sample garment 59 and/or technical pack 54. Further, fabric swatches, or the production sample garment 59 received are/is tested using the fabric testing equipment 122 to produce an initial set of fabric presets, which are tested as described below to produce a final set of presets.
With reference to
In step 350, from the digital pattern 57, production sample garment 59, technical pack 54, grading rules 53, fit model scan data or body measurements 58, and/or paper pattern 51 received from the retailer 50, digital pattern pieces are created, or converted from digital pattern 57, using the apparel product development software 114. Generally, a pattern refers to the collection of the individual pieces of the garment 59. In standard practice, the pattern pieces are drafted first, then laid over fabric, which is then cut around the perimeter of each piece. The resulting pieces of fabric are then sewn together to form the finished garment 59. Therefore, the pattern refers to a blueprint of the garment 49 and its individual pieces.
Indeed, there are several cases in which a digital pattern 57 is received, made, or modified from the above-referenced information received from the retailer 50. In one embodiment, part of the apparel product development software 114 may include a software program named TUKACAD running on product development workstation 116 in the 3D virtual apparel processing system 112, which may be used to create or reformat the digital pattern. TUKACAD is widely used CAD software for digital pattern making, digitizing, grading, and marker making in the apparel industry, and is available from TUKATech, Inc., 5527 E. Slauson Ave., Los Angeles, Calif. 90040, www.tukatech.com. TUKACAD creates points and interpolates splines between points to create a 2D shape or CAD drawing. Additionally, the digital pattern can be graded in TUKACAD to create larger or smaller sizes. Those skilled in the art would recognize that a variety of CAD software programs may be used to perform the functions carried out by TUKACAD.
As noted above, there are several cases regarding the kind of information that is received from a retailer 50 regarding a production sample garment 59 from which the digital pattern pieces are created in TUKACAD. In a first case, a retailer 50 does not have digital pattern 57 or paper pattern 51 for a production sample garment 59. Retailers 50 that do not have patterns 57 or 51 may provide or utilize a widely used technical pack 54 with specifications for how the style is to be made and/or may provide or use a production sample garment 59 for reference. These instructions are then interpreted in 3D virtual apparel processing system 112 to create a digital pattern.
In a likely second case the customer has paper pattern 51 for corresponding to production sample garment 59. Paper pattern 51 may then be digitized or scanned into TUKACAD software using digitizer or pattern scanner 118. As the paper pattern 51 is being digitized, TUKACAD software draws the pattern in digital form resulting in a digital pattern made of digital pattern pieces.
In a likely third case, the retailer 50 has a digital pattern 57 in a third-party format. The digital pattern may then be converted into the format that can be read by the apparel product development software 114 using built-in conversion tools in TUKACAD Software.
In step 352, generally, the physical fabric of a new garment may be tested and simulated to solve for digital fabric presets to be input into apparel product development software 114 for processing. In order to more precisely simulate the behaviour of fabric in a virtual environment, various intrinsic characteristics or parameters that uniquely define real fabric may be determined. The results of those tests may be the fabric presets, which may be entered into a computer model. In some cases, the fabric presets are not independent variables and further testing may be used to arrive at the final fabric presets. In one embodiment, the computer model comprises a three dimensional (3D) virtual software environment.
In one embodiment, software named E-FIT SIMULATOR, also called E-FIT herein, is used as the computer model. E-FIT SIMULATOR is commercially available from TUKAtech, Inc., 5527 E. Slauson Ave., Los Angeles, Calif. 90040, www.tukatech.com, and is built using 3DS MAX'"'"'s SDK. E-FIT, in one embodiment, incorporates cloth simulation plug-in software, CLOTHFX, which is manufactured by Size 8 Software, and is readily available from TurboSquid, Inc., 643 Magazine St., Suite 405, New Orleans, La. 70130, www.turbosquid.com. E-FIT may be used in conjunction with the aforementioned CLOTHFX software to create 3D virtual apparel, including draping on a virtual model and simulating animation in a 3D environment as described below. This combination of software is currently used commonly by designers and engineers for rapid prototyping of apparel design and development.
Generally, some presets are determined by conducting physical tests on one or more swatches of the fabric from production sample garment 59, while other presets also require an additional virtual test, wherein results from the physical test are compared with results from the virtual test in a process of linear regression, which is used to arrive at the final preset value. For example, there may be three fabric presets for stretch-one for warp, one for weft, and one for shear, which may comprise dependent variables that may not be individually solved-for in an isolated test, but rather may require linear regression using all three parameters to find the final presets.
One of the presets tested comprises stretch and shear resistance. An intrinsic property of cloth or fabric is its ability to stretch, which distinguishes it from a normal rigid body. Fabrics can vary in their ability to stretch, and this characteristic can be quantified. In the physical test of the fabric for this characteristic, the fabric assurance by simple testing (FAST) method known to those skilled in the art may be used. Specifically, the known FAST-3 fabric extensibility test may be used. Procedurally, a first sub-test is performed by hanging a swatch vertically. A weight is attached to the swatch, and the change in length due to the force of gravity is measured. The dimension of the swatch that may be tested is typically 15 cm by 15 cm. The direction selected along which to hang the swatch may depend on the direction of the grain-line of the fabric. That direction is typically known as the warp direction. In one embodiment, the test may be performed in the vertical direction (where vertical denotes the direction of gravity) for three specific orientations of the fabric. Those orientations are the directions of warp, weft, and bias. Weft is the direction perpendicular to warp. Bias is the direction that is 45 degrees from the warp and weft directions. The first measurement may be taken in the warp direction. The length of the swatch in the vertical may be, for example, 15 cm, and a weight of, for example, 100 grams may be attached along the bottom of the swatch, and a new length measurement is taken and recorded. The process is repeated for the weft direction. Finally, in the bias direction, the parameter being measured is called shear. For woven fabrics, measurements in the shear direction may also be made using an additional method, similar to the known KES-FB1 tensile/shear testing. For knits, the process may be the same as described above.
A virtual test for stretch and shear is next conducted. Generally, for virtual tests, E-FIT creates a 3D mesh object for the swatch under test, made in the dimension and shape of cloth, which CLOTHFX simulates gravity, collision with itself, and collision with other objects (or itself), to behave in accordance with how physical cloth would behave in a real environment. Therefore, CLOTHFX as applied to a 3D mesh object is accomplished using a set of algorithms based on known computer cloth simulation theory. The CLOTHFX algorithms are based on modelling the 3D mesh object'"'"'s vertices as having mass, and the connections between vertices as springs. In other embodiments, alternative algorithms based on known research can be used to model the mesh as interacting particles. In either case, widely known algorithms in classical dynamics may be used to find the time-varying displacement of each point in the mesh. Such solutions have constants (such as natural frequency, spring constant, mass, etc.) which can be adjusted such that the mesh behaves like any particular fabric. Therefore, before draping, constants which appropriately model the selected fabric are chosen. These constants would be the fabric presets discussed herein. Additional forces that may be modelled may include damping forces, which simulate the effect of friction and air resistance. In the cases of friction and air resistance, the fabric presets found are the coefficient of kinetic friction, coefficient of static friction, and drag coefficient, respectively.
The cloth simulation algorithms used in E-FIT and CLOTHFX are thoroughly described in, for example: Xavier Provot, Deformation Constraints In A mass-Springmodel To Describe Rigid Cloth Behavior, Wayne A. Davis and Przemyslaw Prusinkiewicz, editors, Graphics Interface, pp. 147-154, Canadian Human-Computer Communications Society, 1995; Pascal Volino, Nadia Magnenat-Thalmann, Comparing Efficiency Of Integration Methods For Cloth Simulation, Computer Graphics International, pp. 265-272, July 2001; Kwang-Jin Choi, Hyeong-Seok Ko, Stable But Responsive Cloth, ACM Transactions on Graphics, 21(3), pp. 604-611, July 2002, D. E. Breen, D. H. House, M. J. Wozny. Predicting The Drape Of Woven Cloth Using Interacting Particles. In Computer Graphics (Proceedings of SIGGRAPH 94), Computer Graphics Proceedings, Annual Conference Series, pp. 365-372, Orlando (Florida), July 1994; D. Baraff and A. P. Witkin, Large Steps In Cloth Simulation, Computer Graphics (Proceedings of SIGGRAPH 98), Computer Graphics Proceedings, Annual Conference Series, pp. 43-54, Orlando, Fla., July 1998; and Rony Goldenthal, David Harmon, Raanan Fattal, Michel Bercovier, Eitan Grinspun, Efficient Simulation Of Inextensible Cloth, ACM SIGGRAPH 2007 papers, Aug. 5-9, 2007, San Diego, Calif.
In the vertical test, E-FIT and CLOTHFX may create a 3D mesh of the same dimensions of the physical swatch, then hang it vertically, and attach a virtual weight digitally. CLOTHFX is used to apply cloth simulation algorithms to the 3D mesh. Under the force of gravity, the 3D mesh (now behaving as cloth) is deformed or stretched, and the resultant change in length is measured. The simulation occurs using default values found in the physical tests described above for the stretch/shear resistance preset in all three directions. CLOTHFX applies cloth simulation algorithms to the 3D mesh. In order for CLOTHFX to more precisely model a 3D mesh to behave as a particular fabric, regression analysis is used to solve for the presets by repeating virtual tests and adjusting the presets until the results of the physical and virtual tests match.
Another parameter may comprise bend resistance. This measurement involves the way that fabrics differ from rigid bodies in their ability to bend. The resistance to bend is measured with this parameter. In one embodiment, a physical test uses a known method for assessment of the drape of fabrics. A circular swatch, for example, around 15 cm in diameter, may be draped over a circular rigid body, with a smaller diameter than the swatch, which is propped up by a stand. The setup is situated under a light, such that the resultant folds cast a shadow. This is called a projection of the drape. The projection is then photographed, and the surface area of the projected surface is calculated.
A virtual test for bend resistance may be conducted in similar fashion to the physical test. However, instead of measuring the surface area of the projected image (or shadow from the bends), the mesh is flattened within E-FIT. The resultant area of the flattened mesh may be measured and compared with the surface area measured in the physical test. Using regression analysis, the fabric preset for bend resistance may then be adjusted, and the virtual test may be repeated until the surface areas of both tests match, wherein the resultant fabric preset is the final fabric preset for bend resistance.
Yet two other presets may be kinetic and static friction. Fabric draped on a body can experience damping forces that result from friction with the body'"'"'s surface and friction with itself or with other fabric. A physical test for static friction may be performed by sliding a swatch along a surface, with a known coefficient of static friction. The plane is tilted to find the angle, herein known as the repose angle, at which the swatch begins to slide. The repose angle is used to determine the coefficient of static friction, where the coefficient of static friction equals the tangent of the repose angle for an object sliding down a plane. The coefficient of static friction that results from the physical test may be used as the fabric preset, and no further calculation may be required. Therefore, this value is a direct input into CLOTHFX.
In a physical test for kinetic friction, a method is used in which a constant force is applied to a swatch along a plane to measure the value of the applied force at which the swatch travels at constant velocity. In one embodiment, a string is attached to the swatch, which is pulled along a plane with a known coefficient of kinetic friction. The pull force applied is measured using off-the-shelf instruments for measuring force. The pull force that results in a constant velocity of the swatch along the plane is multiplied by the cosine of the vertical angle of the string used to pull the swatch with respect to the plane. Then, the coefficient of kinetic friction is equal to the force applied multiplied by the cosine of the angle from the plane and then divided by the normal force. The coefficient of kinetic friction may be used as the fabric preset and no further calculation may be required. Therefore, this value may be a direct input into CLOTHFX.
Yet another preset parameter is the surface density of the cloth. A swatch of cloth of the same dimensions can have very different weights, depending on the type of textile used to build the cloth and the density of threads used to weave or knit. In the surface density test, the weight of the cloth is measured. In a physical test, a standard scale is used to measure the weight of a swatch. The weight is divided by the surface area of the swatch to arrive at the surface density. The physical test may be a direct input into CLOTHFX as a fabric preset.
Another preset parameter may be air resistance. Cloth will drape differently depending on the how it falls through a fluid, such as air, and how it reacts with air as it moves in space. When airflow is directed at a cloth, some fraction of the air molecules that make up the airflow will permeate or penetrate the cloth, and some will collide, transferring momentum to the cloth and causing it to move (drag force). The resistance to this drag can vary between fabrics.
In a physical test for air resistance, since the resistance to drag is dependent on the coefficient of drag, and the coefficient of drag will be unique from fabric to fabric, the coefficient of drag is measured. One or more values for the air resistance presets provided by CLOTHFX may be used. However, those skilled in the art would recognize that other well-known tests to measure air resistance could be used to determine such presents for air resistance.
After completing the tests to obtain a final set of fabric presets, the fabric presets 181 may become part of a library of virtual fabrics in the first data storage 110, to be applied when creating virtual apparel made of specific fabric, removing the need to re-test the fabric with new garments made of the same material.
The next step, step 354, comprises preparing digital pattern 180 of the production sample garment 59, either by converting digital pattern 57 from another format, digitizing or scanning paper pattern 51, or creating it using information contained in technical pack 54 Digital pattern 180 may be represented in TUKACAD file format located in data storage 110. TUKACAD'"'"'s file format stores the digital pattern as a collection of points and hermite splines that are interpolated between points. Each point has an attribute that can govern the shape and/or interpolation of the connected hermite splines. Other types of CAD software may use alternative types of splines or interpolation methods, however since all digital patterns can be converted into TUKACAD'"'"'s format, all methods for creating and storing data points in a pattern are supported.
In one embodiment, digital pattern 180 may be made for each particular style in a base size. A base size refers to a sample size of a garment, or size that is used as a standard for a particular garment. Larger and smaller sizes may then be created differentially from this sample size by modifying the digital pattern 180, using a process called grading. The amounts that each point in the pattern are to be moved outward or inward are contained in grading rules 53.
The next step refers to converting the two dimensional pattern pieces into 3D meshes. Once the digital pattern has been prepared, it may be modified with construction information useful for conversion of the 2D pattern into a 3D virtual garment 183. Pattern pieces may need to be adjusted to reduce the complexity of some garment features (e.g., removing extra folds, creating finished pieces for pockets, plackets, etc.). Some values used for physical garment production that are not required for virtual apparel also need to be removed (e.g., fabric shrinkage, sewing allowances, etc.). All of these procedures are made to digital pattern 180 in the TUKACAD software contained in apparel product development software 114. To further explain, the following procedures may or may not be applied to one, more, or all of the pieces of a garment depending on the garment type.
1) First, the digital pattern 180 piece quantity may be adjusted. A few pieces that may otherwise be necessary for production become irrelevant for 3D virtual apparel, and may be removed from the digital pattern 180.
2) Second, sewing allowances may be removed from digital pattern 180. A sewing allowance is an extension of the perimeter of a piece that adds additional fabric necessary for physically sewing a garment. This allowance is not necessary for 3D virtual apparel and may be removed from digital pattern 180.
3) Third, any shrinkage allowance may be removed from digital pattern 180. Digital pattern pieces are often created slightly larger in anticipation that once the fabric is washed, the garment will shrink back to the appropriate dimension. Simulation of shrinkage may not be necessary, and therefore, any allowances for shrinkage in the digital pattern 180 may be removed.
4) Fourth, variable hem lines may be removed from digital pattern 180. Primarily used in men'"'"'s pants, extra fabric is added to the bottom of the pant leg such that a tailor can adjust the hem line. This additional fabric is not necessary for 3D virtual apparel and may be removed from digital pattern 180.
5) Fifth, sewing lines may be added (for pockets, flaps, etc) to digital pattern 180. When a piece needs to be sewn to the inside of another piece, a drill hole may be placed in a physical garment piece. However, in the process of creating digital pattern 180, a sewing line may be drawn digitally to facilitate adding of pockets, flaps, and other features to 3D virtual garment 183.
6) Sixth, a fabric code may be assigned to each piece of the digital pattern 180. For example, the piece that refers to the front of a t-shirt may be assigned fabric code by the name of cotton, whereas the piece that represents the lining of the t-shirt may be given fabric code that represents an elastic material type, such as some polyester spandex blend.
7) Seventh, stitch segments may be assigned in the digital pattern 180. Segments may be defined so that they can be sewn in E-FIT. Marks may be added to the digital pattern 180 to define the starting and ending point of the segments that will be sewn.
8) Eighth, a size may be selected for the fit model avatar 173 (which was created from scan data or measure data from step 58). If digital pattern 180 has been graded into several sizes, the base size may be selected to fit the fit model avatar 173.
9) Ninth, fold lines may be assigned in digital pattern 180. Pieces that are folded (e.g., lapels) may have a line drawn on them where the fold will occur, so that E-FIT can fold the pattern piece along that line.
10) Tenth, pattern pieces may be rotated in digital pattern 180. E-FIT may use the orientation of the pattern pieces as a starting point for making transformations to the 3D mesh. Arranging the digital pattern pieces into a set orientation may ease this process.
11) Eleventh, unnecessary folds may be removed from digital pattern 180. Some pattern pieces may be folded multiple times during the physical construction of the garment. Often, this is not necessary in 3D virtual apparel, and the digital pattern pieces are adjusted to remove this extra length or width from digital pattern 180.
12) Twelfth, internal lines may be adjusted in digital pattern 180. Because the 2D spline pattern pieces are eventually meshed for 3D software, some adjustment of the splines may be necessary to avoid errors in E-FIT. For instance, a line cannot be meshed. So if there is an internal pattern line that extends past the outer boundary of the pattern piece, that external part of the line may need to be removed from digital pattern 180.
The next step 356 may be to convert the digital pattern into a 3D mesh. A 3D mesh, or polygon mesh, is a collection of vertices, edges and faces that defines the shape of a polyhedral object in computer graphics. The mesh is a collection of several closed surfaces. In a mathematical vector algebraic sense, which may be important for calculations, a mesh is a collection of numbers organized into several matrices. More simply stated in a geometric description, a mesh is made of points that are joined together with segments and surfaced by polygons.
In step 356, the digital pattern 180 may now be imported into E-FIT. The CLOTHFX plug-in in E-FIT may convert the pattern pieces into 3D mesh objects. Essentially, the 2D splines are surfaced to create a 3D mesh. The digital pattern 180 is now a 3D mesh. The 3D mesh is then further defined to have components such as pieces and segments, which later get defined with additional attributes.
In step 358, E-FIT interprets the fabric code for each piece of digital pattern 180 and assigns the corresponding fabric presets. For example, the piece of digital pattern 180 that represents front of a t-shirt may have been assigned a material code for cotton. E-FIT interprets this code and retrieves the fabric presets for cotton from its fabric library of presets.
In step 360, E-FIT may apply 3D piece placement, orientation, and curvature in the 3D pattern.
In step 362, E-FIT assigns sewing instructions. In this step, E-FIT matches each particular segment of a 3D mesh corresponding to a particular piece to another segment on the same 3D mesh, or to another 3D piece, in accordance with how the garment is supposed to be sewn together.
With reference back to
Sewing and draping may be completed using functions provided by CLOTHFX and native E-FIT according to the sewing instructions assigned above. Often, garments have lining and/or layers of material. In such cases, layers may be placed, stitched, and draped in a specific order. The culmination of the simulation results in a drape on fit model avatar 173 that may be identical to the drape of a real garment on a real fit model.
With reference to
With reference back to
In step 368 a presentation may be created for the retailer 50 to be approved and later presented to consumer 20. Making an object in 3D appear like physical object may often involve duplicating the look not only in 3D software or interactive rendering software, but require visual output hardware (such as a monitor or display) to accurately replicate the appearance of the object in reference to a real object.
E-FIT may apply a texture. In one embodiment, the 3DS MAX is used as the 3D engine for E-FIT. Since 3DS MAX refers to “textures” as “material textures,” the term “textures” will be referred to as such herein. However, it is understood by those skilled in the art, that the term “texture” is used for an embodiment that does not include using 3DS MAX, but rather some other 3D software, such as PHOTOSHOP available from Adobe Systems Incorporated, 345 Park Avenue, San Jose, Calif. 95110-2704. A material texture 188 contains data that may be assigned to the surface or faces of a 3D mesh so that it appears a certain way when rendered. Material textures 188 affect the color, glossiness, opacity, and the like, of the surface of a 3D mesh.
However, these material textures 188 may not be photometric, in the sense that they may not simulate the interaction of light or photons with the material textures 188 accurately. A user may use E-FIT'"'"'s material editor built-in functions to further create the illusion of the garment'"'"'s appearance. More specifically, the user of E-FIT may work to simulate the correct appearance of material textures by adjusting and applying various material texture properties or texture maps that model the color, roughness, light reflection, opacity, and other visual characteristics.
In one embodiment, material textures 188 may be applied to the surface of each 3D mesh corresponding to each pattern piece. These material textures 188 realistically simulate various attributes that make up the appearance of production sample garment 59. The following list of attributes may be modelled:
- a. color: combination of ambient, diffuse, specular, and/or filter
- b. roughness or bumpiness: bump maps, or displacement maps
- c. light reflection: shiny, glossy, matte, etc which are accomplished using general shader settings or maps.
- d. opacity.
Certain attributes may be set by the retailer. For example, a retailer may send a color swatch with a specific red-green-blue (RGB) value or PANTONE color value. In instances where the appearance is dependent on the lighting conditions, the attributes may be adjusted at the retailer'"'"'s discretion.
Prints, images, logos, and other maps can be adjusted in size, position and orientation. The retailer may provide information (included in technical pack 54) on the placement (position) and size of these maps. Using E-FIT, a user loads these maps and adjusts them accordingly. Furthermore, stitch textures, a component of material texture 188, are added to give the appearance of actual stitching threads.
Completing the above steps results in the completion of 3D virtual garment 183 and fit model drape 186, which are then stored in data storage 110.
Additionally, in step 370, media, such as images, movies, may be rendered and stored as original sample rendered media 182. Additionally, original sample 3D viewer data 187 may be created.
With reference back to
The previous discussion, in section “3D Virtual Apparel”, has been focused on the “3D Virtual Try-On”, a process of draping the existing 3D virtual apparel garment on a consumer avatar is described. Since both processes require the use of an avatar, the following section describes processes to create an avatar, whether the avatar is for a fit model or a consumer.
An avatar may be defined as a 3D mesh constructed to have a similar shape as the consumer body 22 or fit model body 151 it was intended to model, and may or may not be animated. Fit-model avatar 173 may be created to drape 3D virtual garment 183 on the avatar to produce fit model drape 186, by way of system 112. Likewise, consumer avatar object 171 may be used for simulating the drape of production sample garment 59 on a consumer'"'"'s body 22, resulting in consumer drape 1102. The methods for any avatar, whether it be creating consumer avatar 171 or fit model avatar 173, are interchangeable and are described below.
In one embodiment, consumer avatar 171 or fit-model avatar 173 can be generated using three types of procedures, all of which are well-known to one skilled in the art. The first procedure utilizes a technique in which one mesh is conformed to another. The second procedure utilizes a technique called morphing, where one mesh is morphed to another. A third technique involves manually moving vertices from a mesh to another location, which is often called digital 3D sculpting. With respect to creating an avatar, these techniques involve moving vertices from one position to another. However, the conforming and morphing methods are discussed in more detail herein. These two techniques may have disadvantages and advantages over each other and therefore are used in varying situations. Described next is one embodiment of using each of these techniques. However, any technique not discussed, but well known to those skilled in the art could theoretically be used.
An avatar is created using avatar software application 904, which may be contained in avatar processing system 160. Avatar software application 904 begins creating an avatar by first accepting some input data on the consumer or fit-model. There may be many categories of input data, relating to any type of information on a human being or population of human beings—e.g., demographic information. For example, one may have data on the distribution of fat on the human body. Another example is data describing the amount of heat energy emanating from a body. A third example may be the color of the skin, eyes, and hair, and a fourth example may be data on the shape of the body. Since there are many types of information that can describe a human being, it is worthwhile to categorize the information or data. In one embodiment, the following three categories of data may be used to create an avatar: (1) body shape data, (2) body appearance/cosmetic data, and (3) body function data, where body may be defined to include all or any parts of the body, and data may be qualitative and/or quantitative, and stored in any form or format. For example, but not by way of limitation, the term body may include the torso, head, face, hands, fingers, finger nails, skin, hair, organs, bones, etc, or it may only include the torso.
Body shape data, refers to data that can be used or interpreted to understand and reproduce the accurate shape of a human body subject. Body appearance/cosmetic data, refers to data that helps reproduce the appearance of a human subject (e.g. eye color, hair style, skin texture). Body function data provides information on how the human subject'"'"'s body functions. In (e.g. the systems of the body, such as lymphatic, endocrine, skeletal, immune, and others). It may aid to have body function data on movement (e.g. how the body'"'"'s limbs, torso, head, or skeletol, muscular, etc respond to movement). Such data, for example, and not by way of limitation, may be captured using a generic motion capture technology for capturing body movement data. Finally, each data category may have many different types data in which information relating to that category are stored. The various data types for each data category are described below.
Beginning with the first category of data, body shape data, there may be three data types in which information on the shape of a human subject can be stored, provided, or retrieved for use in creating an avatar. For example, but not by way of limitation, the input data may be one or the following: (1) raw body scan data 172, (2) body measurements and other shape data 176 and (3) photographs 174. Although photographs can also be a raw body scan data type, photographs taken in some other mechanism, (e.g. webcam or single camera) may also be included.
Raw body scan data 172 refers to raw output data from any type of scanner, whether it be generic body scanner 149 (e.g. point cloud originating from RF data, structured light data, lasers, mirrors, or any other type of raw data output from these scanners or other yet undiscovered types of scanners.). Moreover, raw body scan data can originate from stereophotogrammetry body scanner 152
Body measurements and other shape data 176 may refer to both manual measurements taken of consumer body 22 either by the consumer or by a third-party, extracted body measurements from raw scan data 172, statistically derived measurements from sizing survey data 178 or avatar statistical data 179, and/or any combination thereof.
Photographs 174 refer to supplemental photographs of the body from different angles, which may or may not include the other parts of the body (e.g. face, hands, etc). For example a user may take a photograph of the face of consumer body 22, and submit the photograph online, by which the system may map the person'"'"'s face to consumer avatar object 171. Photographs 174 may not originate from a scanner, but rather may originate from a web cam, a single digital camera and may be user submitted. Photographs 174 shall not be confused with photographs originating from raw body scan data 172, especially in the case of the method of stereophotogrammetry as described below.
When creating an avatar, the highest precision in reproducing the shape, appearance and function may be desired, however, where precision in data is lacking, a combination of data types may be used to help supplement data or data precision that may be lacking. Therefore, in one embodiment, a combination of data types may be used to further increase the precision of the an avatar.
For example, but not by way of limitation, one may use the following combination of data types for accurately reproducing the body shape of a human subject. These data types could include size survey data. Sizing survey data 178 refers to body measurement and shape data from a population of human beings. For example, but no by way of limitation, the widely used Size USA survey, provided by TC2, which contains raw scan data or extracted body measurements from over 10,000 people can be used. Such data may represent one or many populations with various demographic characteristics. Then, this data may be searchable or queried by a specific demographic or set of demographics. Then, additional information collected on the consumer or fit model such as, age, ethnicity, sex, residence, etc may be used to match the consumer to a specific population that is represented in sizing survey data. If a consumer is matched to a specific population, using demographic data in user data 177, then the body measurements or other shape data for that population may be used in part or in entirety to create the avatar of the consumer or fit model. In yet another embodiment, once a sufficient collection of consumer avatars 171 is gathered, statistics on body measurements and shape can gathered and stored as avatar statistical data 179 and may be used for statistical interpretation and later mined for trends that can further be used to constrain other estimates of the shape of the body, or further enhance those estimates.
Once information, of any data type, regarding the three data categories discussed above, is gathered, the next step is to interpret the data and create an avatar. However, in order to create an avatar, it may be useful to first create one or many base avatars 158. Base avatar 158 is a template avatar from which all other avatars can be made. Depending on the data type for the body shape category of data, the base avatar 158 can be morphed or conformed into the shape of consumer body 22 or fit model body 151
With reference to
In step 382, the base avatar 158 is rigged. A bone structure (or biped) may be inserted into the mesh using 3DS MAX tools, and may be sized and scaled appropriately so that the bone structure fits within the mesh properly. This process is known to those skilled in the art as rigging.
In step 384, within 3DS MAX, the bone structure may be attached to the vertices on base avatar 158 mesh so that when the bones move, base avatar 158 will move in accordance with how a human body typically moves. This process is known to those skilled in the art as skinning, and is not to be confused with putting skin on, which falls into the category of texturing. A file that holds the skinning data may be saved in avatar processing system 160 in avatar data storage 170.
Base avatars 158 can be created for male and females for any typical sample size (i.e., men'"'"'s size 40, women'"'"'s size 8, etc.). From these base avatars 158 made from sample sizes, new avatars can be made in any size and shape.
As discussed earlier, the use of the conforming or morphing techniques is dependent on the type of data received on consumer body 22 or fit model body 151. If the data type is raw scan data 172, then a mesh is created from the raw scan data, and the base avatar 158'"'"'s mesh is conformed to it. In another embodiment, the received data type may be body measurements and other shape data 176. In such a case, the morphing technique may be used. In this case, the base avatar 158 mesh is morphed. The following discussion relates to the case where the data type is raw scan data 172.
Generally, in the prior art, consumer avatar 171, and fit model avatar 173 would be created by measuring the shape of a consumer'"'"'s body, or a physical fit-model described above, by way of a set of measuring tools, such as lasers, cameras, structured light, radio waves, or other electromagnetic based tools. Such configurations of measurement are typically called direct or passive body scanners, and will be collectively referred to as body scanners herein. In one embodiment, stereophotogrammetry system 150 may comprise any of these prior-art types of body scanning technologies, or alternatively, stereophotogrammetry system 150 may include stereophotogrammetry body scan booth 152 described below. Stereophotogrammetry system 150 may also comprise any body scanning software for processing raw scan data to create 3D meshes or avatars. Alternatively, stereophotogrammetry system 150 may include body scanning software 154 described below. For example, companies that produce some of these types of prior art scanners include those available from Unique, 133 Troop Avenue, Dartmouth, NS, B3B 2A7, Canada, TC2/Imagetwin, located at 5651 Dillard Dr., Cary, N.C. 27518, Telmat Industrie, 6, rue de l'"'"'Industrie—B. P. 130—Soultz, 68503 GUEBWILLER Cedex (France), and, or Human Solutions, GmbH, Europaallee 10, 67657 Kaiserslautern, Germany.
However, in one embodiment of the presently described system, stereophotogrammetry may be applied. Photogrammetry is the practice of determining the geometric properties of objects from photographic images. In the simplest example, the distance between two points that lie on a plane parallel to the photographic image plane can be determined by measuring their distance on the image, if the scale of the image is known.
A more sophisticated technique, called stereophotogrammetry, involves estimating the three-dimensional coordinates of points on an object. These are determined by measurements made in two or more photographic images taken from different positions. Common points are identified on each image. A line of sight (or ray) can be constructed from the camera location to the point on the object. It is the intersection of these rays (triangulation) that determines the three-dimensional location of the point. More sophisticated algorithms can exploit other information about the scene that is known a priori, for example symmetries, in some cases allowing reconstructions of 3D coordinates from only one camera position.
Algorithms for photogrammetry typically express the problem as that of minimizing the sum of the squares of a set of errors. This minimization is known as bundle adjustment and is often performed using the Levenberg-Marquardt algorithm.
The stereophotogrammetry method may have advantages in cost and features that other methods cannot achieve. With reference to
Furthermore, the refined photographs are then sent as inputs to 3DSOM PRO software available from About Creative Dimension Software, Ltd., Wey Court West, Union Road, Farnham, Surrey GU9 7PT, United Kingdom. This software then uses these photographs to create 3D mesh 159. However, those skilled in the art would recognize that many different software applications may perform similar functions. 3D mesh 159, is then imported into 3DS MAX, wherein the base avatar 158 is morphed to the dimensions and shape of 3D mesh 159.
With reference to
In step 400, the camera 800 is assembled. Any standard charge coupled device (CCD) or complementary metal-oxide-semiconductor (CMOS) camera 800 can be used. In one embodiment, a CMOS 2 megapixel chip is used in order to maximize resolution while minimizing cost, such as that provided in the QUICKCAM 600 available from Logitech, Inc., 6505 Kaiser Dr., Fremont, Calif. 94555 USA. However, any CCD or CMOS commercially available digital camera, webcam, professional camera, industrial camera, or security camera could be used. The aforementioned QUICKCAM 600 has a 2 Megapixel sized CMOS chip providing 30 frames/second over a universal serial bus (USB) 2.0 connection. The camera 800 may be dissembled to retrieve only the circuit board with the CMOS chip attached and USB still connected. However, any megapixel size chip with any frame rate and other connections (e.g., Firewire), could also be used. Moreover, additional cameras could be added, a slightly rotating pedestal could be used, and/or mirrors could be used in place of some cameras. However, the method described herein was selected due to accuracy and cost-effectiveness.
In step 402, a wide angle lens may be attached to a spacer, attached to a camera enclosure, which encloses the circuit board to which the CMOS chip is attached. A wide field-of-view lens may be used in this embodiment so that the camera 800 can be positioned as close to the consumer body 22 or fit model body 151 as possible while keeping the subject within the field of view. Any distortion due to the lens may be corrected-for in 3D SOM PRO software using its lens calibration tools. In one embodiment, a 2.9-8.2 mm lens, provided by Computar, Inc., 55 Mall Drive, Commack, N.Y. 11725, may be used.
In step 404, a plastic project enclosure (for example, 3×2×1 inches), provided by RadioShack, Inc., may be used to house the camera 800. A 3-5 mm hole may then be cut open to make the CMOS chip visible. A 5 mm spacer with threads may be attached over the hole and the lens is screwed into the spacer.
Steps 400-404 may be repeated for each camera to be used.
In step 406, stereophotogrammetry body scan booth 152 is assembled. Standard zero structures 910 may be used to assemble the structure, for example, a 7 ft×7 ft×7 ft sized stereophotogrammetry body scan booth 152. A matte 920 with a specific pattern, which may be provided by 3D SOM, Inc., may be placed in the center of the floor 915. This is where the consumer body 22 or fit model body 151 stands. Cameras 800 and lights may be fixed to cross beams 912 that attach to the four pillars of the structure 910 along the perimeter. Electrical pipe may be built around the structure on the inside and outside of the zero pillars at the very top of the body scanning booth 152. Fabric may be hooked to the pipes to create drapes to enclose the structure from outside light, and to include a fixed color background behind the subject from all angles. Pre-fabricated structures could be used in a similar manner, where modifications may be made depending on the type of structure.
Referring again to
In step 410, the cameras 800 are connected in an array. Cameras 800 may be connected USB powered hubs in one embodiment. All hubs may be connected to a computer with USB ports. In other embodiments, the cameras may be wired for Bluetooth, Ethernet, wifi, or the like.
In one embodiment, stereophotogrammetry body scanning software 154, which may interface with or include software components, may also contain executable instructions to perform one or more of the following steps 412-418 described below. In step 412, the video stream of consumer body 22 or fit model body 151 is acquired. MATLAB software, which may be one of the software components of stereophotogrammetry body scanning software 154, is available from Mathworks, Inc., 3 Apple Hill Drive, Natick, Mass. 01760-2098, and which may be used to read the video streams from the cameras. Specifically, the image acquisition toolbox of MATLAB may be used to start and view all 20 video streams. Those skilled in the art would recognize that a variety of software programs may be used to perform the functions carried out by MATLAB.
In step 414, the images may be acquired from the video stream, wherein the main subject is consumer body 22 or fit model body 151, and may be placed in the middle of the stereophotogrammetry body scan booth 152 to stand on a matte, such that their body is in the field view of the cameras. The cameras are triggered to acquire images or single frames from each camera 800. In one embodiment, a manual trigger may be used with cameras that do not support hardware triggering. However, hardware triggering can be used to speed up image acquisition to prevent any lag time between cameras.
In step 416, MATLAB'"'"'s image processing toolbox may be used to mask images, save them in any format that can be read by 3D SOM PRO, and send them to 3D SOM PRO Software. Software written using MATLAB may be compiled into a standalone executable file to perform this step.
In step 418, 3D mesh 159 is created using 3D SOM'"'"'s software.
In one embodiment, the number of cameras 800 may be arbitrary. By way of example, and not by way of limitation, 20 or more, or less, cameras 800 may be used. Further, the position of the cameras 800 may be more or less arbitrary in one embodiment. A position calibration map 820 may be used for helping the 3D SOM PRO software determine the position of the cameras 800 in three dimensional space. In one embodiment, the position calibration map 820 may comprise a flat annular component having radial spaced black circles 822 printed thereon. Depending on the position of each camera 800, the black circles 822 are captured by each camera 800 with a different distortion, which 3D SOM PRO, or other software used to calibration position, is capable of interpreting to indicate the position of each camera 800. In one embodiment, the black circles 822 may preferably be of varying sizes.
Further, any number of various types of cameras 800 or sensors may be used. In one embodiment, webcams may be used because they are less expensive and may provide relatively higher resolution with CMOS sensors at the same price. However, more expensive digital cameras with CCD sensors with a broader color ranges may be used. Further, any type of lens may be used with the cameras 800. For example, the lenses are capable of having various focal lengths. For example, the types of lenses may be defined by variations in focal length, diameter, and/or magnification.
In order to calibrate the cameras for such variations in lens types, for example, a lens calibration map 830 having black circles 832 similar to those on the position calibration map 820 may be used. Each camera 800 may be calibrated for type of lens by pointing each camera at the lens calibration map 830 at a constant distance to and angle, taking pictures at various zooms. The 3D SOM PRO software may then use the varying images captured by each of the cameras 800 and/or lens types. The 3D SOM PRO software then takes the calibration images and correct for the varying cameras 800 and/or lens types.
With the above description of the stereophotogrammetry system 152, those of skill in the art would recognize that the stereophotogrammetry system 152, may comprise an arbitrary number of two or more cameras 800 for taking independent photographs of a physical object; a position calibration map 820 for providing three dimensional position data for the two or more cameras 800; each camera 800 having a lens, wherein each lens has a type, wherein two or more of the lenses are capable of being the same type; a lens calibration map 830 for each type of lens, wherein the lens calibration map is capable of correcting for non-linearity within the lens; a first set of instructions capable of execution on a processor 153 to acquire a set video streams from the two or more cameras 800; a second set of instructions capable of execution on a processor 153 to trigger the two or more cameras 800 substantially simultaneously to produce an image from each camera 800; a third set of instructions capable of execution on a processor 154 to download and save the image from each camera 800; a fourth set of instructions capable of execution on a processor 153 to mask the image from each camera 800 to produce a set of masked images; a fifth set of instructions capable of execution on a processor 153 to process three dimensional positional data from the position calibration for the set of masked images; and a sixth set of instructions capable of execution on a processor 153 to process a three dimensional mesh from the set of one or more masked images. The system 153 may have a variable number of cameras 800. The system 152 may include variable positions of the cameras 800. The position calibration map 820 may be modifiable according the number and position of the cameras 800. Further, the lens calibration map 830 may be modifiable according the types of lenses on the cameras 800. The size of the whole stereophotogrammetry system 154 may also be adjustable. The first, second, third and fourth software instructions may also comprise image acquisition and processing software instructions, which may all be embodied in the body scanner software 154. The image acquisition and processing software instructions may comprise MATLAB software instructions in one embodiment. The image acquisition and processing software instructions may comprise LABVIEW software instructions in another embodiment.
In one embodiment, the download of the images from the cameras 800 may occur using universal serial bus (USB), Firewire or wifi network devices.
The fifth and sixth software instructions may comprise three dimensional modelling software. In one embodiment, the three dimensional modelling software may comprise 3DSOM PRO. In another embodiment, the three dimensional modelling software may comprise compiled object oriented software instructions.
Lights 840 may be a part of the system 152, which may be used to create uniform lighting conditions to create the least amount of shadows. Reflectors may be used to further achieve ambient light conditions within the booth 152. A uniform background may be used within the walls of the booth to aid in the masking process. Those skilled in the art, for example, may find a green background generally aids in the masking process.
Finally, the size of the stereophotogrammetry body scan booth 152 may be variable or adjustable, generally having little effect on the operation of the booth 152. This allows for the booth 152 to be adjusted for use in different special arrangements as space may provide.
With reference to
In another embodiment, a prior art body scanner system 149 may be used in place of stereophotogrammetry system 150, where prior art body scanner 149 may refer to all currently existing forms of body scanners described in prior art, or alternatively all other body scanners contemplated by future technologies. Then, prior art body scanner system 149 may also provide a 3D mesh as an output. In this case, the initial step performed by avatar software application 904 is step 427, similarly importing the 3D mesh 159.
However, in another embodiment, output data from prior-art body scanner 149 may only provide raw scan data as input in step 425, and not a 3D mesh. Thus, in step 426 3D mesh 159 may be created from a prior-art scanner system'"'"'s 149 raw scan data using MESHLAB software, a widely available open source application available form http://meshlab.sourceforge.net/, 3DS MAX, and/or any 3D software able to perform such function with raw scan data.
In step 426, 3D mesh 159 is imported in to 3DS MAX software.
In step 428, scaling and alignment of 3D mesh 159 with base avatar 158 may take place. Within 3DS MAX, the base avatar 158 may be superimposed on top of the 3D mesh 159. The base avatar 158 may then be scaled in size such that its height aligns with the height of the 3D mesh 159. When scaling up and down, the shape and proportion of the base avatar 158 may not change. In other words, the system grows or shrinks base avatar 158 so that 3D mesh 159 and base avatar 158 occupy a similar volume. Furthermore, the limbs of base avatar 158 may also be adjusted to align with the limbs from 3D mesh 159.
In step 430, the head, hands, and feet are detached from base avatar 158 in order to complete the next step.
In step 432, the torso of base avatar 158 is conformed to the torso of 3D mesh 159. MAXSCRIPT code, which is a scripting language provided by 3DS MAX, may be run, which can run within 3DS MAX. This script moves verticices of the torso of base avatar 158 to the torso of 3D mesh 159, such that their shapes and proportions are the same and they occupy the same volume. In running this script, the skinning may be lost and can be reproduced.
In step 434, the hands, feet and head of base avatar 158 are re-attached to newly conformed mesh.
In step 436, the conformed mesh is re-skinned using saved data stored in avatar data storage 170.
In step 438, animation is applied. This step may be to store a standard point-cache file which stores the animation components of consumer avatar 171 or fit model avatar 173.
If the subject was consumer body 22 then the conformed mesh may be referred to now as consumer avatar 171. Otherwise, if the subject was fit model body 151 then the conformed mesh may be referred to now as fit model avatar 173.
In step 440, consumer avatar 171 or fit model avatar 173 is exported from 3DS MAX and stored in avatar data storage 170.
In one embodiment, consumer avatar 171 or fit model avatar 173 may be derived directly from body measurements 176 instead of 3D mesh 159, where body measurements and other shape data 176 may have been extracted from raw scan data 172, or from user data 177 (e.g. demographics) using avatar software application 904. Further quantitative information may include data originated from statistical analysis of historical body scans (sizing survey data 178) and/or avatar statistical data 179. If the consumer provides these measurements, they may do so by entering on computing device 24 which then stores the in user data 177. The computing device 24 may comprise any type of processing device, such as a personal computer (desktop or laptop), smartphone, iPHONE®, iPAD®, tablet pc, mobile computing device, kiosk, gaming device, media center (at home or elsewhere), or the like. For example, but not by way of limitation, the consumer may enter body measurements and/or select other avatars features using an html form or a client-side software application 28 running on computer device 24. The user'"'"'s selection and entered data is then to ASP 100'"'"'s avatar software application 904 running in avatar processing system 160.
With reference to
Since the data type is body measurements and other shape data, base avatar 158 may be morphed to create the shape of consumer avatar 171 or fit model avatar 173.
One skilled in the art would recognize that in order to morph a mesh, one may require morph targets. Therefore, base avatars 158 may have morph targets, allowing them to be morphed. For extremely large and small human bodies, additional base avatars 158 may created with additional morph targets. A morph (sometimes called a control) is applied to the base avatar 158 that links to the morph target, and can be used to interpolate between the two objects, changing the size/shape of the base object to match the morph target'"'"'s geometry either partially or completely. In other words, by adjusting the morph target, one can approximate the shape of a new avatar. When several morphs are adjusted such that the new avatar similarly match the consumer body 22'"'"'s or fit model body 151'"'"'s body shape and or measurements, then one has arrived at consumer avatar 171 or fit model avatar 173 respectively.
Each morph target may correspond to one or many points of measure. Points of measure are control points for a specific body measurement from body measurements and other shape data 176 (e.g. the circumferential waist measurement may have a control point). Therefore, when the point of measure needs to be changed to a specific body measurement value (given by the user, extracted from raw scan data, or derived by some other means), the morph target is adjusted.
With reference to
In step 502, base avatar 158 may be morphed as described above.
Another embodiment includes supplementing body measurement 176, user data 177, sizing survey data 178, or avatar statistical data 179 with digital images 174. Digital images 174 from a single camera may further enhance the process of creating consumer avatar 171 or fit model avatar 173. Multiple digital photographs may be used as references for sculpting the mesh of base avatar 158 within avatar software application 904, wherein sculpting refers to the process of adjusting the morph targets to match a visual contour of consumer body 22 or fit model body 151 given in a digital photograph.
With reference to
In step 514, the base avatar mesh is sculpted. The digital photographs may be used as references to match the shape of the avatar to the real person. The photographs may then be mapped to planes in a 3D scene in 3DS MAX and placed around the base avatar'"'"'s mesh. This makes it possible to use the photographs as references to the shape of the body that is being reproduced digitally. For example, if the photograph is front-facing, then the base avatar'"'"'s mesh is also front-facing in the scene. Second, the base avatar'"'"'s morph targets are adjusted to get the shape close to where it should be to match the silhouette of the reference image. Then, vertices in the base avatar'"'"'s mesh are adjusted using soft selection methods to correct the avatar to match the references, and the measurements. When using photographs as references, photographs of the front, side and back of the body are adjusted digitally to correct errors in the photography as much as possible.
In yet another embodiment, the above methods described with respect to creating a consumer avatar 171 may be mixed, matched, and/or combined. For example, body measurements 176 can be further enhanced by adding images from a single camera of the body and face of consumer body 22 or fit model body 151.
With reference to
The next process may include draping the 3D virtual garment 183 on a consumer avatar 171 in an automated process on the web or computing device 24, resulting in consumer drape 1102. The process begins when the consumer chooses to virtually try-on 3D virtual garment 183. The consumer can request to virtually try-on 3D virtual garment 183 by way of a graphical user interface (GUI) on computing device 24, or by sending a request over the internet through a website.
In one embodiment, the consumer may send a request on the internet to virtually try-on a garment by clicking hyperlink 81 which may reside in retailer'"'"'s online store 80, a third-party online store, or on an online store running ASP 100. Hyperlink 81 may be positioned next to a display of a 3D virtual garment 183, or a digital representation of production sample garment 59 available for virtual fitting. When a user presses hyperlink 81 using computing device 24, a sequence of events is started. With reference to
In another embodiment, referring to step 600, a user may run 3D draping software application 900 locally on computing device 24 enabling the user to virtually try on garments. This embodiment may require the user to sign in and exchange data with ASP 100 or retailer system 3D draping software application 900 may run computer device 24 or may run online in ASP 100 as an online service for retailers or consumers over a wide area network through a network connection. 3D virtual try-on processing system 1200 may exist at the retailer or may be hosted by a third party web server. In another embodiment, 3D draping software application 900 may run on kiosk 130. The user may click on a link or a button with a mouse, or interact with a touch screen on the display of computer device 131. The user may see the resultant output of the 3D virtual try-on process on 3D viewer application 132.
In step 604, it is determined whether the appropriate size for the consumer has already been determined. If so, processing moves to step 614. Otherwise, processing moves to step 608, to conduct size prediction algorithm 908.
In step 608, consumer'"'"'s body measurements and other shape data 176 are queried from avatar processing system 160 and compared against 3D virtual garment measurements 184 of 3D virtual garment 183 at corresponding points of measure. The root mean square (rms) of the deviations of these two sets of measurements (body measurements 176 vs. 3D virtual garment measurements 184) is calculated for each size available for production sample garment 59. Ease added to digital pattern 180, may be added to the shape of the avatar to better assist in attaining a solution.
In step 610, it is determined whether the size that results in the lowest rms is sufficient for an initial guess. Those skilled in the art of statistical analysis may use chi-squared or other statistical tests to assess the strength of the initial guess which may depend on the accuracy of which the consumer avatar 161 accurately duplicates the size, shape and proportion of consumer body 22. Moreover the user may determine if the initial guess is sufficient. If it is determined that the size is sufficient to serve as the initial guess for draping, then processing moves to step 614 wherein the initial guess of the 3D virtual garment 183 is queued for draping on the consumer avatar 161. Otherwise, processing moves to step 612 wherein multiple sizes of 3D virtual garment 183 are queued for draping on the consumer avatar 161.
In both steps 612 and 614, queue simulation request(s) is/are performed. Once received, simulation requests are sent to a queue system 903 that is capable of maintaining lists of multiple simulation requests from multiple users.
It is also possible that the user may want to virtual try-on one garment with one or more other garments that either they have previously tried on. If the user has selected to try on multiple garments, step 618, then processing moves to step 620 where the system retrieves consumer drape 1102 that corresponds to the garment that the user wishes already display on their avatar before draping additional clothing.
In step 622, associated files for the simulation that are queued are then retrieved from data storages 110 and 170. For example, all or any combination of files stored in data storages 110 and 170 may be retrieved which may be required for the size algorithm, the simulation and the fit analysis described above.
In step 624, node polling system 912 is initiated. When the simulation request is read and all file locations have been verified, in step 626, the software running the queue system 903 checks the node polling system 912 to find an available GPU 1002. In one embodiment, GPU 1002 may reside in a GPU cloud computing center 1000.
In step 628, the polling system 912 is updated to reflect that the selected GPU 1002 is in use for the simulation request and not available for other simulations.
In step 630, 3D draping software application 900 then continues by processing the simulation on the selected GPU 1002.
The 3D draping software application 900 may be EFIT with slight modifications. For example, but not by way of limitation, 3D draping software application 900 may run EFIT without a GUI and user action. In other words, in one embodiment, 3D draping software application 900 is simply EFIT software that has been modified to run automatically by accepting simulation requests from the queue, loading the appropriate files, processing the simulation by draping the garment on one or more CPUs or GPUs, and then exporting the required output files
Processing involves draping 3D virtual garment 183 on consumer avatar 161. The existing fit model drape 186 on fit model avatar 173 may be loaded onto consumer avatar 161. Then, the drape process may be continued to readjust to account for the difference in the two avatars. The resultant output is consumer drape 1102. Processing of cloth simulations in a 3D environment may be hardware-intensive. To those skilled in the art, GPUs 1002 are preferred for simulation of 3D graphics. However, when GPUs 1002 are not available, more traditional CPUs may be used in their place. In one embodiment, GPUs 1002 or CPUs can be run in parallel to increase simulation processing speed through multi-threading so long as the selected processor supports it.
Moreover, processing may include simulating for animation. In such a case, an animation file is loaded. The animation file may be of consumer avatar 161 walking, running, dancing, sitting, or performing any human motion. Draping is performed on each frame of animation of consumer avatar 161 and then stored in consumer drape 1102.
With reference to
Focusing back to
In step 636, the node polling system 912 is updated to reflect that the selected GPU 1002 is now available.
In step 638, a fit analysis algorithm 906 may executed in order to determine qualitative and quantitative data with respect to the outcome of the simulation (the 3D virtual try-on process). A fit analysis object may be created to store this qualitative and quantitative data. The output of fit analysis algorithm 906 may also be fit data 1104 and/or rendered media 1108. Fit analysis may include deriving qualitative and quantitative data from a consumer drape 1102 for multiple sizes for a specific garment, or just one single size.
Fit analysis algorithm 906 may perform a stretch test to determine how much the virtual fabric is stretching in consumer drape 1102. Positive stretch values may indicate tighter fit areas, zero or a small stretch value may indicate areas of good fit or simply no-stretch. Negative stretch values may indicate areas of compression. In one embodiment, stretch values may be used to determine how well or how poor a garment fits an avatar. This data can then be stored additionally as fit data 1104.
Stretch can be calculated in many ways. For example, but not by way of limitation, stretch may be calculated by measuring the percent difference in a specific measurement before and after the drape. In other words, an initial garment measurement might yield one length. After draping the garment on an avatar, the draped garment measurement at the same location might have a length that has increased or decreased. In one embodiment, the percent difference in length for that specific measurement may be defined as the stretch value. In another embodiment, the stretch value may be calculated for many garment measurements, and the stretch value may refer to the total stretch of all garment measurements, or the average stretch value of all garment measurements.
Quantitative data may also include calculating the change in stretch in a similar fashion as described above, but with initial value set to the stretch value of the base size, and the final value being the stretch value of the selected size (if other than the base size). Furthermore, quantitative data may also include calculating the stretch value for specific points of measure, rather than for the entire garment, and then comparing them with the initial 3D virtual garment measurements from fit model drape 186. Moreover, quantitative data may also include calculating the total volume of space between the garment and the body and assessing how that total volume may increase or decrease from size to size. All data may be used together, or in pieces in a decision engine to establish a prediction of size. The decision engine may consider the total volume between the garment and the body, from size to size, versus the total stretch value, from size to size, and weight the two data types to arrive at the best fit of the garment to the body. It is well known to those skilled in the art that common procedures are available to determine how a garment is fitting using specific points of measure.
With reference to
With reference to
Yet another form of visual representation of the consumer drape can be replacing the existing material texture of 3D virtual garment 183 with a one inch by one inch grid pattern, which is applied as a material texture, which reveals the slope or curvature of the garment along the body. Fit analysis algorithm 906 may perform many other types of calculations. For example, but not by way of limitation, fit analysis algorithm 906 may calculate the total volume of space, using methods in calculus, between 3D virtual garment 183 and consumer avatar 161 for all sizes of consumer drape 1102. This volume may aid in interpreting the correct size of the garment. Moreover, this calculation may aid in interpreting the fit of a garment.
The data gathered from the fit analysis algorithm, whether it be quantitative or qualitative or both, stored as fit data 1104, becomes extremely useful information to retailer system 50 and consumer system 50. More about this fit data will be discussed later
Referring back to
In step 642, 3D viewer data 1112 and fit data 1104 are displayed in 3D viewer application 82 or 132. 3D viewer application 82 may be embedded in webpage viewed on browser 26 or is an application on consumer computing device 24. In another embodiment, 3D viewer application may run in ASP 100 and may be viewable in browser 26.
In one embodiment, 3D viewing application 82 or 132 is an interactive renderer java applet made with Java and Java 3D libraries, each available from Oracle/Sun, 500 Oracle Parkway, Redwood Shores, Calif. 94065, with built-in functionality to rotate, pan, zoom, and animate virtual garment 183 on consumer avatar 171. The user may also view the drape of one size larger or smaller than the estimated size. The user can also select to view the current virtual garment 183 with a color tension map, in x-ray mode, playback animation of the drape, or view the garment with the avatar hidden from view. Moreover, the user can render an image to save in common image formats. 3D viewer application 82 or 132 may also have other interactive features that allow the user to rotate, pan, and zoom the 3D content. The user may also be able to annotate the garment with comments. Moreover, live sharing and chatting may be implemented so that the user can share the content live with another user. Chatting and video applications may be embedded allowing users to communicate further and discuss the 3D content.
Discussed above was an embodiment of 3D viewer application 82 or 132 written in Java and Java 3D. However, it is important to note that 3D viewer application 82 may be an interactive renderer created using c++, python, or any programming language capable of creating 3D web applications.
In one embodiment, in step 644, the user can rate and/or review the fit of the garment by giving a thumbs-up or thumbs-down. In another embodiment, the user can rate and/or review the garment on a numeric scale. In yet another embodiment, the user can rate the garment as “Fits well, too tight or too loose”. Other rating systems known to those skilled in the art can be used. All such reviews described above can be stored in 3D virtual try-on data storage 1100 as user reviews 1106.
In step 646, the user are given the option of saving consumer drape 1102 of 3D virtual garment 183 for future viewing or mixing with other garments for viewing (e.g., shirt and pants). If saved, virtual garment 183 appears in user'"'"'s virtual closet 290 where the collection of consumer drapes 1102 are available for the user to view again. The user'"'"'s subsequent action(s) are tracked within the application and/or webpage to determine whether they purchase the garment. If the user chooses to purchase the garment, an email notification may automatically be generated to the user notifying them that the virtual garment 183 has been saved in their user profile and can be viewed at any time by logging into the ASP 100'"'"'s web portal using computing device 24.
Virtual closet 290 may be accessed when the user is logged into ASP 100. Virtual closet 290 may store consumer drapes 1102 of 3D virtual garments 183 that have been purchased and recently viewed. In one embodiment, virtual closet 290 may display these garments 183 as visual images of drapes that do not include the model.
Items in the closet may be viewed in 3D viewing application 30 can be viewed with other 3D virtual garments 183, for example, from the same retailer, or a different retailer, or mixed and matched in other ways.
In some embodiments, the virtual closet 290 may also provide for sharing between users. With social media integration, a user may share the results of their fit with contacts in facebook, myspace, yelp, and other social media sites, as well as personal websites or for viewing in applications in any computing device. The user may select a save image function that allows the user to take a picture or snap shot of the consumer drape 1102 of 3D virtual garment 183 on the avatar, and then upload it to their profile on a social media site.
With the data collection (consumer drape 1102, fit data 1104, user reviews 1106, rendered media 1108, and consumer avatar 171) that is accomplished by system 10 described herein, such data may be analyzed to discover trends and draw conclusions, which can, for example, provide feedback into the system and provide further granular analysis (step 306 in
Then, in one embodiment, in step 704, the data can be mined to find trends in buying behaviour, trends in consumer drapes from one garment to another, and or trends in body shapes with particular garments or particular retailers. For example, but not way of limitation, stretch factor calculations for relevant points of measure calculated for the virtual garment 183 could be analyzed across multiple garments for a single user, or multiple users.
Moreover, in step 704, trends in stretch factor, or other fit data may be correlated with demographics, retailer'"'"'s, fit model'"'"'s, sizes, fabric types, revealing valuable information. For example, but not by way of limitation, such analysis may reveal that a consumer fits better with a certain set of brands, then with another set of brands. Such information becomes useful in step 706. Moreover, such correlations may be easily recognized by those skilled in the art given the data the present system makes available, since brands often have fit models with distinctively different body shapes.
In step 706, the trends discovered in step 704 may be used to better predict the outcome of fits with virtual garments in system 10 and can be used as size prediction algorithm 908. Furthermore, fit may be a very subjective personal choice for consumers. For instance, two people of very similar body types may have dramatically different viewpoints on fit, where one person may prefer a tighter fit, or a size larger than the other. Therefore, by studying how variables that measure stretch across multiple garments for groups of similar bodies, and discovering trends, those trends may now be applied to predict other garments that may fit a user.
In step 708, a product recommendation engine is built to interpret predicted garments in step 706 and then suggest those garments to the user in ASP 100.
Finally, data collected can be used directly to make custom patterns and therefore custom garments for the consumer. The data may be used to develop block patterns, or customize the patterns of garments available by the retailer. Custom 3D garments and patterns may be sent to the retailer based on the analysis.
Conversely, consumer drape 1102, fit data 1104, user reviews 1106, and rendered media 1108 may all contain extremely valuable information not only for aiding consumers in buying clothing online, but also for apparel manufacturers and retailers. Retailers can use such information to better understand their target market, make necessary adjustments to product development, distribution, production, merchandising, and other key decisions in supply chain and sales processes referred to above. Currently, retailers have no immediate perceivable method of determining how a garment truly fits on each of their customers. Often times, retailers depend on statistical studies to determine the body shape(s) of their target market. Moreover, they rely on third-party research organizations that study body shapes in certain populations. However, the shapes of human bodies are difficult to standardize and are constantly changing. In consequence, most retailers fall short in reaching the broad target market they were designing for.
With reference to
In step 744, a web interface, may be made available to retailers. By logging on, retailers may have access to daily, weekly, monthly, quarterly, or yearly statistics on user data, which can be manipulated and searched.
Range Cameras may include, for example, the Microsoft 3D Kinect device. With reference to
In one embodiment, the system described herein may use the depth images attained by the CMOS sensor 2604 to create a 3D model of a subject or object within the field of view. Further, a process of capturing depth images of a human subject and creating a 3D model or avatar of the subject may be performed by one embodiment.
With reference to
Each depth image may contain the depth or distance to the body, as well as the xy position of each part of their body, also called 3D position data.
In step 2701, a library routine of OpenNI™ may be called to calculate actual 3D points from the captured depth images from step 2700. In step 2702, consumer body 22 may next be rotated or rotate to a secondary position, by way of example, and not by way of limitation, 90 degrees.
Next, in step 2704, a second series of one or more images may be captured in a second interval of time. In step 2705, the library routine of OpenNI™ may be called to calculate actual 3D points from the captured depth images from step 2704.
The process is repeated until the subject has rotated 360 degrees, as indicated by decision diamond 2706. The result is a series of 3D points, one set for each capture of images at a rotation stop point as described above.
In step 2708, each set of 3D points corresponding to a rotation of the consumer body 22 is rotated and translated such that they all are able to fit together to form a final set of 3D points to represent the entire consumer body 22. This final set of 3D points are stored in step 2710.
Next, in step 2712, measurements may be extracted. This may be performed various convex-hull algorithms, for example, the Graham scan algorithm or the Andrews monotone convex-hull algorithm.
In step 2714, a 3D mesh is created from the 3D points. This can be performed by various methods that are commonly used to convert 3D points to a 3D mesh. For example, ball pivoting algorithms, Poisson surface reconstruction, or the like, may be used for this step.
In step 2716, the mesh may be converted into 3D consumer avatar 171 as described above. For example, the mesh could be rigged, skinned, and have a texture applied so that it could be animated and customized to look like the consumer body 22. In step 2722, the consumer 22 could then use this consumer avatar 171 for an online fitting room as described above. As described above, clothing could be modelled as a 3D mesh, as in the case with digital patterns, and then using the cloth simulation algorithms described above, clothing may be simulated on the avatar in 3D, allowing for the consumer 171 to view in real-time how a garment will look and fit their own body.
In some embodiments, another sensor could be put behind the consumer 22, or several at different angles. However, to keep hardware cost down and to make the system more practical for in-home use, consumer 22 may alternatively be asked to rotate their body to capture their body from multiple angles as described above.
In step 2714, the corrections in change of posture may be made by using a pose tracker library by OpenNI. The OpenNI library contains functions for tracking poses by assigning a skeleton to the consumer body 22. For example, if the arm position has changed from the first series of images, to the next series of images after the body was rotated, then using the pose tracker, the new position of the arm can be used to translate the 3D points associated with the arm to the old position of the arm in 3D space, thereby, correcting for movement by the user.
Alternatively, the consumer avatar 171 could also be drawn on a monitor or flat-panel display connected to a computer or gaming system, and then be synced with the consumer'"'"'s movements, such that the consumer could control its movements.
Using a technique known as augmented reality, one skilled in the art of augmented reality systems would recognize that 3D graphics could be displayed on a live video stream from RGB camera 2606. Those 3D graphics could be consumer avatar 171.
3D virtual garment 183 draped on consumer avatar 171 could also be displayed using augmented reality and dynamically draped using GPU cloth simulation. In this respect, 3D virtual garment 183 may be simulated with animation in real time on consumer avatar 171 no matter what position or posture consumer avatar 171 takes in real time.
Moreover, consumer avatar 171 could be hidden from view such that it would appear to the consumer 22 that the 3D virtual garment 183 were actually on consumer body 22 as they see it in real time on the monitor.
For example, consumer body 22 may change poster wherein the arm may change position in 3D space, using the pose tracking algorithm developed in OpenNI™, consumer avatar 171 may adjust its position to match the new position of consumer body 22. Since the consumer avatar 171 hidden, this will thus cause 3D virtual garment 183 to re-simulate using the cloth simulation algorithm resulting in a new drape consistent with consumer body 22'"'"'s new posture.
Inexpensive depth sensor technology, sometimes called range camera technology, has become readily available by multiple vendors for consumer and commercial applications. Moreover, open source software libraries and software development kits (SDK) have been provided by those vendors to develop new applications utilizing this technology. Presented here is a new application that utilizes multiple range cameras to create a 3D human full or partial body scanner (hereafter 3D body scanning system). In one embodiment, the 3D body scanning system is mounted on a pole or several poles, vertically or horizontally, or mounted on a wall or several walls within a booth or within a room of any size. Additionally, the sensors may be angled. The sensors are configured to scan, via software, either all at once, within some timing of each other, or in some pre-defined order, to capture depth and 3D position data of a subject. An associated RGB camera may also be used to assign a color for each 3D position data point (a collection of such 3D position data points hereafter referred to as a “point cloud”). The resultant point clouds, with or without color information, can then be registered or transformed into one common coordinate system, via a registration algorithm, creating one point cloud representing a full subject or parts of a subject. Then, that point cloud may be interpreted by a meshing algorithm, using standard practices or a novel method, to create a triangular or polygon mesh (hereinafter mesh). Finally, that mesh can be used for a virtual try-on (as described above). Described below are multiple sensor position configurations, multiple registration algorithms and calibration assemblies, multiple sensor timing and ordering schemes used for capturing depth images, and multiple meshing algorithms used for converting the resultant 3D point cloud into a mesh.
Range cameras may comprise, by way of example and not by limitation, devices available from PrimeSense, 28 Habarzel St., Tel-Aviv, 69710, Israel, the Xtion Pro or Xtion Pro Live from ASUS Computer International, 800 Corporate Way, Fremont, Calif. 94539, the Kinect available from the Microsoft Corporation of Redmond, Wash., or devices from other vendors that use range sensors or cameras. With reference to
With reference to
Body Scanning System Configuration
The following describes multiple relative placements of range cameras around a subject, hereinafter referred to as body scanning system configuration or range camera configuration.
In several preferred embodiments, one or more cameras may each be positioned along two or more vertical axes surrounding the scan area. By way of example and not by limitation, with reference back to
Several constraints may limit the range camera placement described above. Some of those constraints are as follows: (1) the volume available for the 3D body scanning system (by way of example and not by limitation, with reference back to
With reference to
Sensor Timing and Ordering
Sensor timing is affected by several constraints, including range camera frames per second and range noise/accuracy. Inaccurate range measurements may in some embodiments necessitate taking the average or taking the median of depth or RGB values of multiple range images from a particular camera or multiple cameras. Additionally, it is important to note that increased accuracy and precision comes at a cost of longer scan time, and a longer scan time increases the risk of a human subject, in one embodiment, changing their posture or moving their body. By way of example and not by limitation, one embodiment captures 10 frames per range camera, but other embodiments may capture fewer or greater number of frames per range camera. In another embodiment, the number of frames captured by each range camera could vary, depending on the camera.
One consequence of using multiple range cameras that rely on projected patterns to acquire range data is that different cameras may interfere with each other if their fields of view overlap each other, resulting in one or more range cameras attempting to capture data of the same region at the same time, resulting in interference and therefore missing range data. This necessitates a scan order, where individual range cameras or groups of range cameras are turned on and off sequentially or within some timing of each other. Because range camera interference is dependent on relative placement to each other, the timing and ordering scheme used for an embodiment may be dependent on each range camera'"'"'s position and line of sight relative to each other.
Additionally, turning on/off the range cameras may also require an amount of time. Therefore, in embodiments that utilize a scan order, the group of cameras first in the scan order is turned on prior to a scan of a subject. These cameras may also aid in positioning of the subject in the scan area.
By way of example and not by limitation, with reference to
Range Camera Registration
The following section describes multiple methods for registering each range camera'"'"'s point cloud into a common coordinate system, which may be one of the range camera'"'"'s coordinate systems or any other coordinate system, effectively aligning together the point clouds from each range camera. Registration into a common coordinate system may be required to generate a complete point cloud of a subject.
In one embodiment, each range camera'"'"'s point clouds are registered via an iterative closest point (ICP) algorithm, known to those skilled in the art of machine vision. An initial estimate of the parameters required to register a range camera point cloud to another, hereinafter referred to as transformation, may be determined by the range camera configuration described above. Each pair-wise transformation may then be refined using the ICP algorithm, and the refined transformation data may be saved for future body scans. ICP may be used, for example, where there is sufficient overlap between each camera'"'"'s point cloud. ICP may also be used in combination with other methods. ICP is described in detail in A Method for Registration of 3-D Shapes by Besl & McKay, which appears in IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 14, no. 2, pp 239-256, February 1992.
In other embodiments, one or more objects, possibly in a predefined configuration. By way of example and not by way of limitation, the objects could be a triangular arrangement. The predefined configuration will be referred to herein as a calibration assembly, which may be used to facilitate registration. With reference to
In embodiments where a calibration assembly is used, a calibration procedure may be required prior to scanning a subject. With reference to
With reference to
With reference back to
In some embodiments, a combination of shapes and other objects may be used in a calibration assembly. Some embodiments may use poles, rods, strings, or some other method for holding the calibration objects from the bottom, from the sides, or some other position. Some embodiments may use a combination of calibration assemblies and methods described above.
With reference to
With reference to
One or more calibration points and/or vectors may then be extracted from each object'"'"'s point cloud. With reference to
In step 4910, the vector from the origin of range camera 2600 to point 3120 may then be extended by the radius estimate, which yields a coarse sphere center estimate 4912. In step 4914, this estimate 4912 may then be used as an initial guess to a nonlinear least squares least fit algorithm, known to those skilled in the art of computer graphics, to attain a more accurate estimate of the sphere center 4916.
In reference back to
In reference back to
In reference back to
In one embodiment, a method based on Poisson surface reconstruction (PSR), known to those skilled in the art of computer graphics, is used to, with reference to
In step 5306, in some embodiments, a scalar field f(P) is utilized, which may or may not be the same as that used in PSR. In one embodiment, the scalar field is defined such that a point (not necessarily part of the point cloud) inside the body is assigned a different value than a point outside the body, and thus calculating the value of the scalar field at a point reduces to determining whether a point is inside the body. One embodiment uses the concept of shadow volume, known by those skilled in the art of computer graphics, with each range camera analogous to a light source and the range camera'"'"'s line of sight analogous to light rays.
In step 5302, in one embodiment, due to, for instance, missing depth values in the raw 2D range images, perspective differences between cameras, and inconsistent range/pixel resolution, the 2D range images used for determining shadow volume of each range camera are constructed from points originating from multiple cameras. In another embodiment, the aforementioned range images are constructed from the entire point cloud of the subject. In either embodiment, 2D range image construction is performed by transforming points into each camera'"'"'s 2D projective coordinate system and, for each pixel, using the smallest range value.
In step 5304, in one embodiment, to obtain a set of oriented points, also known as an oriented point cloud, which may be used as input to PSR, the point cloud of the subject is inserted into an octree (also oct-tree) data structure, known to those skilled in the art of computer graphics. The octree assigns a rectangular prism volume to each data point. A detailed description of the octree is contained within Oct-trees and their use in representing three-dimensional objects by Jakins & Tanimoto, published in Computer Graphics and Image Processing, Volume 14, Issue 3, November 1980, pp 249-270. Then, in step 5306, f(P) for each of the eight vertex points of the rectangular prism are determined, after which, in step 5308, the marching cubes algorithm, known to those skilled in the art of computer graphics, may be used to triangulate within the rectangular prism. A detailed description of the marching cubes algorithm is contained within Marching cubes: A high resolution 3D surface construction algorithm by Lorensen & Cline, published in ACM SIGGRAPH Computer Graphics, Volume 21, Issue 4, July 1987, pp 163-169. Then, in step 5310, every centroid of each triangle face is retained along with its corresponding normal vector, defined as the normal to the triangle face in a consistent direction. In one embodiment, this direction is inside the body, while in another embodiment, this direction is outside the body. The retained oriented point cloud 5312, used as input to PSR 5314, may contain a different set of points than the original point cloud 5300 of the subject. The output of PSR is a triangular mesh 5316. In other embodiments, the marching cubes algorithm alone may generate a triangular mesh.
In some embodiments, parts of the body surface may be sparsely represented by points in the point cloud of the subject and/or there may be gaps in the point cloud.
In one embodiment, the oriented point cloud determined by marching cubes may be more dense than required. To save on processing time during PSR, this oriented point cloud may be thinned out based on a radius threshold. That is, for each oriented point, other oriented points within a distance threshold may be removed from the oriented point cloud.
In one embodiment, it may be desirable to smooth the surface of the triangular mesh created by PSR. By way of example and not by way of limitation, Laplacian smoothing, known to those skilled in the art of computer graphics, may be applied.
In one embodiment, since the processing of a node in an octree is independent of the processing of any sibling node, the above steps may be parallelized. This may result in significant runtime efficiency.
In one embodiment, elements of the preceding preprocessing steps may be merged with PSR. This can be done since PSR uses an octree data structure. By way of example and not by way of limitation, calculating f(P), using marching cubes triangulation to define surface normals, and/or splitting octree nodes during marching cubes may be performed on the same octree used by PSR.
The meshing method used in any embodiment may use elements from a combination of the embodiments described above.
Most apparel produced and purchased today is built with standard grading or sizes conceived by the retailer, brand, or manufacturer. That sizing was pre-selected by the retailer, brand, or manufacturer using internal or published data on body shapes in various markets.
However, body scanning and subsequent measurement extraction paves the way for retailers, brands, and manufacturers to create a custom size for an individual. That custom size essentially reflects either a new draft of a digital or paper pattern that reflects the consumers body measurements and additional ease or fit requirements set by the consumer or retailer, brand, and manufacturer. In other cases, the custom size is simply an alteration of an existing size, that again, reflects the body measurements of the consumer.
The process of adjusting or automatically drafting a new customized digital pattern uses techniques that one skilled in the art of pattern making would easily recognize.
Additionally, a consumer may customize the decoration or design of a garment by selecting colors, textures, prints, and other artwork and placing it on any part of the garment. A user interface that runs on multiple platforms may provide the consumer the ability to customize the design of the apparel. The design process may occur on a 2-dimensional digital pattern displayed in the user interface, or a 3D rendering of the garment draped on an avatar. The avatar may be of the consumer with a simulated drape of the garment in a chosen or recommended size or a default of an avatar. If the design process occurred in 3D, using a 3D renderer or 3D images, the final design may then be translated back into the 2D geometry of the pattern including process-related elements necessary for garment production, including sewing allowance, bleed area for dying and other elements, and then be sent to a manufacturer for production.
The manufacturer may dye, print, and cut, in any order, depending on the technology the manufacturer has adopted, and then sew the final garment for delivery.
Automatic spreaders or manual spreaders may be used to lay fabric in preparation for cutting each individual piece of the customized garment.
Automatic cutting technology may be used to automatically cut each individual piece of the customized a garment. In another embodiment, manual cutters may be used in place of automatic cutting machines.
In another embodiment, wherein the customization is centered on only the design or decoration of the garment, but not the adjustment or automatic creation of the pattern, pre-cut pieces in various sizes may already be produced and kept as inventory.
Waterless printing and chemical free dyeing technology may be used to automatically print and dye the design of each piece already cut, or cut specifically for the consumer.
In another embodiment, pre sewn garments may be used and designs may be placed on the pre-sewn garments, often called blanks, using screen printing or sublimation printing technology.
The user interface may display the final design to the consumer and then enable them to purchase a garment before production takes place. In the case that automatic printing and dyeing technology is used to print and dye on-demand on each piece, and the pieces are cut on-demand, a retailer, brand and manufacturer could keep zero inventory, and thus, wait for an order or transaction from the consumer before beginning production.
With reference to
With reference to
With reference to
Over and above the mobile phones and tablets 5880, computers 5860, and the like, discussed above, other popular devices, such as modern game consoles 5870, are now capable of video play. Game consoles 5870 such as the XBOX®, Playstation®, Nintendo®, Wii®, and others, provide for internet video presentation. Just as with the IPTV box 5892, game consoles 5870 typically connect to a TV 90 on which videos may be viewed and games played.
One or more servers 5840 may include one or more storage devices 5848 containing one or more databases. One or more scanning systems 2600 with having a range camera 2800 is further connected to the internet as described above in more detail.
With reference to
With reference to
With reference to
With reference to
With reference to
In step 6304, all new remote records are downloaded from the server 5840 from database 58250. In step 6306, the system iterates through the downloaded records. In step 6318, the system checks for whether each record exists locally. If so, then the remote record is merged into the local record in step 6320. If not, then in step 6321, the local record is created.
In step 6308, the system iterates over the local scan files. For each file, the system checks for whether a local record exist in step 6322. If so, then, in step 6324, the information from each scan file is loaded into a local record, which is created in step 6326.
With reference to
The present invention is a 3D body scanner, that scans a full body quickly, for example in less than a minute, creates a 3D representation of the body, sometimes called an avatar, and then displays that avatar to a user. The avatar is then further measured by a digital tape measure and then those digital measurements are used to determine specific biomarkers for health, wellness, fitness, clothing recommendations, and more. A software application drives the body scanner, processes the images for the body scanner, and displays the scanned body. In an alternative embodiment, the operation of the scanner, the processing of the resultant data, and the displaying of the data in an interface to the consumer could all exist on different computing devices or every combination thereof. The need for 3D body scanning technology, particularly in the fitness industry is strong. Within the fitness industry, people are often misled to believe that their total weight is the primary metric to track progress. But often times, weight can fluctuate unpredictably during a fitness program or through dieting as the body'"'"'s composition is transforming. The resultant total weight could remain relatively unchanged, leading the person to believe that he or she is not making progress. This source of confusion then discourages people from continuing with their plan and may result in them quitting altogether. Because fat takes up more volume than muscle, the body actually shrinks when it loses fat. The resultant shrinkage can be measured via a tape measure, but can be difficult to track with manual measurements, given their susceptibility to human error. A 3D scan creates a more objective view of the body and can be compared against future scans to isolate changes in the shape of the body due to fat loss. Furthermore, the digital tape measure is significantly more precise, and small changes in waist, hip, and other circumference measurements are more readily seen.
Alternative embodiments may include a built-in monitor that illustrates the correct arm position, via a live depth feed, or via a playback animation, or a video. The turntable 820 is further enhanced to include a digital weight scale, with, for example, eight load cells positioned beneath the surface of the platform. By using a turntable 820 that duals as a weight scale, users need not enter their weight manually as with other systems. The present embodiment also adds a bluetooth or other wireless radio 6802 (
The tower 6500 contains angled depth cameras or sensors 2600 that enable a wider view of the arms, when positioned in an A-pose. These angled sensors 2600 therefore allow for further reduction in the footprint. Examples of the positioning of the sensors 2600 is shown in
With reference to
With reference to
In one embodiment, the plates 6700 are mounted to a rail 6704 that is welded to a half cylindrical aluminium tube that makes up the body of the scanner tower 6500. The tube may be cut to almost entirely in half, leaving a bottom portion intact. In the bottom portion, that remains a full cylinder, a mini-PC may embedded that comprises the processor 6508. The sensors 2600 may be attached via a universal serial bus (USB) hub 6012 to the processor 6508. Through the USB hub 6012, depending on the available ports resident in the processor 6508, the speaker 6504, LED lights 6502, and bluetooth or wireless radio 6514 may be attached to USB ports on the processor 6508.
With reference to
With reference to
In one embodiment, the system software on the embedded PC 6508 uses the captured depth images to help recognize features on the body that can be interpreted as a gesture or the correct or incorrect position of the arms or feet. In one embodiment, the captured images are then processed to create a 3D mesh and used to extract measurements as described above.
The configuration of the presently described embodiment reduces the footprint of a previous embodiment by more than 50%, and as a result, enables smaller businesses with less space the ability to adopt the solution. Besides adding multiple range camera'"'"'s 2600, a further reduction in the footprint is achieved by embedding the processor 6508 within the enclosure. Additionally, keeping both the turntable 820 and scanner tower 6500 separate allows for flexibility in the footprint. For example, a shorter subject may not need to be positioned on a turntable that is 22.5 inches from the base of the sensor tower. Because their body is smaller, the tower 6500 could be placed closer, further reducing the footprint. Previous integrated body scanning systems were forced to accommodate the widest and tallest person which meant they wasted space for smaller individuals. Additionally, the system uses LED lights, a voice assistant, and gesture to automate the scanning procedure. These elements enable the system to be intuitive to set up and use. Moreover, by creating a bluetooth connection from the turntable 820 and the sensor tower 6500 and embedding the PC 6508 to the tower 6500, a further reduction in space was achieved, as well as an enhancement in the ease of use. The present embodiment allows any computer device to wirelessly operate the system.
In addition to the fitness industry, there exists a strong need for precise body measurements within the clothing industry. The present invention provides surface measurements of the body, such as circumferences which could help tailors make better fitting clothing, brands and manufacturers produce better sizing, and retailers recommend size and fit for each customer. Moreover, the medical field could benefit from a non-invasive and accurate way to measure body composition using the highly accurate and precise surface body measurements produced by the scanner. Early pilot studies have shown that the measurements produced by the present invention are highly correlated with body composition measurements measured by medical gold standards. Academic, 3D printing, Gaming, and other fields could make use of the body measurements or 3D avatar.
The various embodiments described above are provided by way of illustration only and should not be construed to limit the invention. Those skilled in the art will readily recognize various modifications and changes that may be made to the claimed invention without following the example embodiments and applications illustrated and described herein, and without departing from the true spirit and scope of the claimed invention, which is set forth in the following claims.