Apparatus for producing 3D point-cloud model of physical object and producing method thereof

• US 10,495,459 B2
• Filed: 01/03/2018
• Issued: 12/03/2019
• Est. Priority Date: 10/30/2017
• Status: Active Grant

• Current Assignee: Kinpo Electronics Incorporated (Cal-Comp Electronics Public Company Limited), XYZprinting, Inc.
• Original Assignee: XYZprinting, Inc.
• Inventors: Chen, Wei-Fan, Qureshi, Waqar Shahid
• Primary Examiner(s): Liu, Gordon G
• Application Number:
  US15/861,660
• Publication Number:
  US 20190128670A1
• Time in Patent Office: 699 Days
• Field of Search: 345420
• CPC Class Codes: G01B 11/24 for measuring contours or curvatures
• CPC Class Codes: G01C 11/025 by scanning the object
• CPC Class Codes: G06T 17/00 Three dimensional [3D] modelling, e.g. data description of 3D objects
• CPC Class Codes: G06T 2207/10028 Range image; Depth image; 3D point clouds
• CPC Class Codes: G06T 7/337 involving reference images or patches
• CPC Class Codes: G06T 7/529 from texture
• CPC Class Codes: G06T 7/55 from multiple images
• CPC Class Codes: G06T 7/579 from motion

Abstract

An apparatus for producing 3D point-cloud model of physical objects includes a rotatable platform (11), multiple patterns asymmetrically arranged on the rotatable platform (11), a background curtain (13), and an image capturing unit (12) arranged facing toward the background curtain (13). A producing method includes: placing an object (2) to be scanned on the rotatable platform (11); setting a capturing amount during one-time rotation operation; controlling the rotatable platform (11) to perform the rotation operation, and controlling the image capturing unit (12) to capture corresponding images during the rotation operation according to the capturing amount, wherein each image includes the entire object (2) and multiple of the patterns, and records corresponding global coordinates; and, performing matching on multiple features of each of the images, and constructing a 3D point-cloud model of the object (2) according to a matching result of the features and the global coordinates of each of the images.

Claims

1. An apparatus (1) for producing 3D point-cloud model of physical object, comprising:

a rotatable platform (11), arranged to carry an object (2) to be scanned;

multiple patterns, arranged on the rotatable platform (11), wherein the multiple patterns are asymmetric patterns or being arranged on the rotatable platform in an asymmetric way;

a background curtain (13), arranged at one side of the rotatable platform (11);

an image capturing unit (12), arranged at other side of the rotatable platform (11) against the background curtain (13) and aiming at the background curtain (13); and

a microprocessor (10), electrically connected with the rotatable platform (11) and the image capturing unit (12), configured to set a capturing amount of the image capturing unit (12) during a one-time rotation operation of the rotatable platform (11);

wherein, the microprocessor (10) is configured to receive a trigger action for controlling the rotatable platform (11) to perform the one-time rotation operation and controlling the image capturing unit (12) to capture multiple images of the object (2) according to the capturing amount during the rotation operation, wherein each of the multiple images respectively records the entire object (2) and part of the multiple patterns, and respectively records a set of global coordinates corresponding to different angle of view, wherein each of the multiple images respectively has a content and an angle of view different from other images of the multiple images and the different part comes from the object (2) or part of the multiple patterns;

wherein, the microprocessor (10) is configured to perform a matching action on multiple feature points of the object (2) and the multiple patterns of each of the multiple images, and re-builds a 3D point-cloud model of the object (2) according to a matching result of the matching action and the sets of global coordinates of each of the multiple images, and the re-built 3D point-cloud model contains at least a set of data of 3D point-cloud of the object (2) and another set of data of 3D point-cloud of the multiple patterns.

2. The apparatus (1) in claim 1, further comprising a tablecloth (5) arranged on a surface of the rotatable platform (11), wherein the tablecloth (5) is recorded with the multiple patterns thereon.

3. The apparatus (1) in claim 1, wherein the rotation operation is to rotate 360 degrees in a constant speed, the multiple images have same capturing intervals, and the multiple images respectively have different contents and different angles of views.

4. The apparatus (1) in claim 1, further comprising a storage (14), electrically connected with the microprocessor (10), wherein the storage (10) is configured to record the multiple images, the multiple feature points, and the 3D point-cloud model.

5. The apparatus (1) in claim 1, wherein the microprocessor (10) is configured to retrieve the multiple feature points from each of the images, to perform the matching action on the multiple feature points of each two images of the multiple images, and to select two images of the multiple images having most matched feature points as an initial pair of an incremental reconstruction procedure.

6. The apparatus (1) in claim 5, wherein the microprocessor (10) is configured to perform a multipoint least error of regression calculation on the matched feature points of the two images being the initial pair for obtaining a partial 3D point-cloud model, wherein the partial 3D point-cloud model is configured to record a plurality of data of a 3D point-cloud within a tolerance scope, and the microprocessor (10) is configured to continually add multiple feature points of a non-processed image of the multiple images to the partial 3D point-cloud model one-by-one, and to continually perform the multipoint least error of regression calculation to adjust and increase the data of the 3D point-cloud of the partial 3D point-cloud model, so as to obtain a complete 3D point-cloud model.

7. The apparatus (1) in claim 1, further comprising a second image capturing unit (15), electrically connected with the microprocessor (10) and arranged aiming at the background curtain (13), wherein the second image capturing unit (15) is arranged with a height and an angle different from that of the image capturing unit (12), and the second image capturing unit (15) is set with the capturing amount same as the image capturing unit (12).

8. A method for producing 3D point-cloud model of physical object, adopted by a producing apparatus (1) having a rotatable platform (11), multiple patterns, a background curtain (13), an image capturing unit (12), and a microprocessor (10), the rotatable platform (11) arranged to carry an object (2) to be scanned, the multiple patterns arranged on the rotatable platform (11), the background curtain (13) arranged at one side of the rotatable platform (11), the image capturing unit (12) arranged at other side of the rotatable platform (11) against the background curtain (13) and aiming at the background curtain (13), the microprocessor (10) electrically connected with the rotatable platform (11) and the image capturing unit (12), and the method comprising:

a) setting a capturing amount of the image capturing unit (12) during a one-time rotation operation of the rotatable platform (11) by the microprocessor (10);

b) receiving a trigger action by the microprocessor (10) to control the rotatable platform (11) to perform the one-time rotation operation and to control the image capturing unit (12) to capture multiple images of the object (2) according to the capturing amount during the rotation operation, wherein each of the multiple images respectively records the entire object (2) and part of the multiple patterns, and respectively records a set of global coordinates corresponding to different angle of view, wherein the multiple patterns are asymmetric patterns or being arranged on the rotatable platform in an asymmetric way, each of the multiple images respectively has a content and an angle of view different from other images of the multiple images and the different part comes from the object (2) or part of the multiple patterns; and

c) performing a matching action on multiple feature points of the object (2) and the multiple patterns of each of the multiple images respectively by the microprocessor (10), and re-building a 3D point-cloud model of the object (2) according to a matching result of the matching action and the sets of global coordinates of each of the multiple images, wherein the 3D point-cloud model contains at least a set of data of 3D point-cloud of the object (2) and another set of data of 3D point-cloud of the multiple patterns.

9. The method in claim 8, wherein the producing apparatus (1) further comprises a tablecloth (5) arranged on a surface of the rotatable platform (11), and the tablecloth (5) is recorded with the multiple patterns thereon.

10. The method in claim 8, wherein the step b) is to control the rotatable platform (11) to rotate 360 degrees in a constant speed, and to control the image capturing unit (12) to capture the multiple images according to same capturing intervals, wherein the multiple images respectively have different contents and different angles of views.

11. The method in claim 8, wherein the step c) comprises following steps:

c1) retrieving the multiple feature points of each of the multiple images respectively;

c2) performing the matching action on the multiple feature points of each two images of the multiple images;

c3) selecting two images of the multiple images that have most matched feature points as an initial pair of an incremental reconstruction procedure after step c2);

c4) executing a multipoint least error of regression calculation on the multiple matched feature points of the two images being the initial pair for obtaining a partial 3D point-cloud model, wherein the partial 3D point-cloud model is configured to record a plurality of data of a 3D point-cloud within a tolerance scope;

c5) determining whether the multiple images are all processed completely; and

c6) before the multiple images are all processed completely, continually adding the multiple feature points of a non-processed image of the multiple images to the partial 3D point-cloud model one-by-one, and re-executing step c4) to adjust and increase the data of the 3D point-cloud of the partial 3D point-cloud model, so as to obtain a complete 3D point-cloud model.

12. The method in claim 8, further comprising a step a0) before step a):

automatically determining and adjusting a light source and a light strength of the producing apparatus (1) suitable for the object (2) to be scanned.

13. The method in claim 8, further comprising a step c0) before step c):

performing a filtering process on each of the multiple images for deleting one or more feature points that are not belonging to the object (2) and the multiple patterns from each of the images.

14. The method in claim 8, further comprising a step d):

outputting the 3D point-cloud model for executing a modeling procedure of a virtual 3D object corresponding to the object (2).

15. The method in claim 8, wherein the producing apparatus (1) further comprises a second image capturing unit (15), electrically connected with the microprocessor (10) and arranged aiming at the background curtain (13), wherein the second image capturing unit (15) is arranged with a height and an angle different from that of the image capturing unit (12), and the second image capturing unit (15) is set with the capturing amount same as the image capturing unit (12), and the method further comprises following steps before step a):

a1) controlling the rotatable platform (11) to perform the rotation operation and controlling the image capturing unit (12) and the second image capturing unit (15) to capture an alignment picture on the rotatable platform (11) for generating multiple alignment images during the rotation operation, wherein each of the multiple alignment images respectively contains the entire alignment picture;

a2) retrieving multiple records of intersection position information of each of the multiple alignment images respectively;

a3) performing a matching action on the multiple records of intersection position information of each two alignment images of the multiple alignment images;

a4) performing a regression calculation according to a matching result of the matching action for estimating relative positions between the image capturing unit (12) and the second image capturing unit (15);

a5) calculating an average height difference according to a result of the regression calculation for obtaining a height difference of the image capturing unit (12) and the second image capturing unit (15);

a6) applying the result of the regression calculation to a circle equation for estimating a rotating radius; and

a7) utilizing a center of the rotatable platform (11) as a zero point and estimating actual relative positions of the image capturing unit (12) and the second image capturing unit (15) according to the height difference and the rotating radius.