BODY MEASUREMENT

• US 20130315470A1
• Filed: 08/08/2012
• Published: 11/28/2013
• Est. Priority Date: 05/25/2012
• Status: Active Grant

• Current Assignee: Quantacorp Nv (BodiData, Inc.)
• Original Assignee: Poikos Limited
• Inventors: Watson, Eleanor, Evans, David
• Granted Patent: US 8,842,906 B2
• Time in Patent Office: Days
• US Class Current: 382/154
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06T 2207/10004 Still image; Photographic image
• CPC Class Codes: G06T 2207/20081 Training; Learning
• CPC Class Codes: G06T 2207/30196 Human being; Person
• CPC Class Codes: G06T 7/75 involving models

Abstract

A method of generating three dimensional body data of a subject is described. The method includes capturing one or more images of the subject using a digital imaging device and generating three dimensional body data of the subject based on the one or more images.

Claims

1. A method of generating three dimensional body data of a subject;

said method comprising;

receiving one or more images of the subject from a digital imaging device;

partitioning the one or more images into a plurality of segments;

analysing the segments of the one or more images to determine the probability that the subject is located in the segment;

identifying one or more distributions within the or each partitioned image, each distribution relating to a property of the one or more images;

utilising the probabilities and the distributions to produce one or more unique probability maps representing the subject;

comparing the one or more unique probability maps with a database of representations of three dimensional bodies to determine a best mapping between the or each unique probability map and a representation determined from the database; and

generating three dimensional body data of the subject based on the best mapping.

2. The method of claim 1, wherein identifying one or more distributions within each partitioned image, each distribution relating to a property of the one or more images further comprises:

weighting pixels in the or each image according to the determined probability to generate a corresponding weighted image, and

identifying a colour and/or spatial distribution within these weighted images

3. The method of claim 1, further comprising:

providing one or more probability map(s) that identify the expected shape and pose of the subject in the one or more images.

4. The method of claim 1, further comprising:

determining the pose of the subject and providing one or more probability maps based on the determined pose.

5. The method of claim 1, wherein comparing the one or more unique probability maps with a database of three dimensional body representations to determine a best mapping comprises:

generating an average three dimensional body image from the database of three dimensional body representations;

generating at least one silhouette of the average three dimensional body image; and

comparing the silhouette with the one or more unique probability maps of the subject and determining the degree of match.

6. The method of claim 5, further comprising:

providing a predefined series of vectors that define how the average three dimensional body image can be altered;

applying the predefined series of vectors to the average three dimensional body image to generate a modified three dimensional body image;

generating a modified silhouette of the modified three dimensional body image; and

comparing the modified silhouette with the unique probability map of the subject and determining the degree of match.

7. The method of claim 6, further comprising:

comparing the degree of match of the silhouette and the degree of match of the modified silhouettes to determine the best match silhouette.

8. The method of claim 7, further comprising:

altering the best match silhouette to account for imaging factors present within the one or more images of the subject.

9. The method of claim 8, wherein the imaging factors present are at least one of:

camera pitch;

camera yaw;

camera distance and barrel distortion.

10. The method of claim 7, further comprising:

determining the relative measurements of the best match silhouette based on a LOOKUP table of known three dimensional body images measurements; and

scaling the best match silhouette based on reference data provided by the subject, said reference data being one or more of;

subject height;

distance between subject and camera;

reference to a calibrated known object.

11. The method of claim 3, wherein analysing the segments of the one or more images to determine the probability that the subject is located in the segment comprises:

comparing the location of pixels in the or each image with their corresponding locations on the probability maps to determine the probability that a pixel represents the subject in the image.

12. The method of claim 1, wherein after utilising the probabilities and the distributions to produce one or more unique probability maps representing the user, the method further comprises:

blurring the edges of the one or more unique probability maps to compensate for variable line sharpness to produce a boundary of uncertain pixels;

identifying the gradient of the edges of the one or more unique probability maps to determine the direction of the subject at each uncertain pixel to produce a corresponding vectored outline probability map; and

applying the vectored outline probability map to the image to determine points with the highest contrast; and

identifying, at said point of highest contrast, a boundary pixel located on the edge of the subject.

13. The method of claim 12, further comprising:

comparing the colour of the boundary pixel with the colours of neighbouring pixels to determine from the intensities a true boundary position of the subject with sub-pixel resolution.

14. The method of claim 13, wherein the true boundary is calculated using:

b=1−

(c_B−

c_A)/(c_C−

c_A), where b is an adjustment value added to the boundary pixel to determine the true boundary position of the subject, c_B is the intensity of the current pixel, c_A is the intensity of the previous pixel along the vector and c_C is the intensity of the next pixel along the vector.

15. The method of claim 11, wherein a colour distribution is identified within the weighted images and identifying the colour distribution additionally comprises:

sampling pixels of the image to identify their colour; and

mapping the location of the sampled pixels to the pre-rendered probability map to determine the expected colour of that region of the probability map.

16. The method of claim 15, wherein identifying the colour distribution further comprises:

obtaining a probability distribution of the colour of the pixels within all said segments of the image; and

generating a three dimensional (multivariate) probability distribution of the image.

17. The method of claim 15, further comprising:

comparing the probability distribution with the expected colour of corresponding regions of the probability map to determine the probability that a pixel represents the subject.

18. The method of claim 1, wherein combining the probabilities comprises calculating the unique probability that a pixel represents the subject using the equation:

19. The method of claim 1, further comprising:

converting each of the one or more images into an RGB image for identifying defects in the one or more images.

20. The method of claim 1, further comprising:

outputting the three dimensional body data of the subject.

21. The method of claim 20, wherein the three dimensional body data is output to one or more of:

a storage device;

online cloud based storage;

a mobile phone;

a display screen;

or a personal computing device.