Object pose recognition

Patent metadata

Field	Value
Publication number	US-9818195-B2
Application number	US-201615074438-A
Country	US
Kind code	B2
Filing date	Mar 18, 2016
Priority date	Mar 20, 2015
Publication date	Nov 14, 2017
Grant date	Nov 14, 2017

Abstract

A method for use in estimating a pose of an imaged object comprises identifying candidate elements of an atlas that correspond to pixels in an image of the object, forming pairs of candidate elements, and comparing the distance between the members of each pair and with the distance between the corresponding pixels.

First claim

The invention claimed is:

1. A computer implemented method for use in estimating a pose of an imaged object, the method comprising the following steps:
a) receiving an image made up of image elements, the image representing, in three spatial dimensions, a scene including the imaged object;
b) for each of a plurality of the image elements, identifying one or more corresponding candidate locations in an atlas of one or more candidate objects;
c) forming a pair of image elements from the plurality of image elements, the pair comprising a first image element and a second image element;
d) determining a first distance between the first image element and the second image element;
e) determining a second distance between a first candidate location corresponding to the first image element and a second candidate location corresponding to the second image element; and
f) comparing the first and second distances.

2. The method of claim 1 , further comprising performing steps (c) to (f) for one or more further pairs of image elements from the plurality of image elements.

3. The method of claim 2 , further comprising the step of g):
i) selecting a first pair of candidate locations having first and second candidate locations corresponding to the first and second image elements of a first of the pairs of image elements; and
ii) selecting a second pair of candidate locations having first and second candidate locations corresponding to the first and second image elements of a second of the pairs of image elements,
iii) forming a triplet of candidate locations, the triplet comprising the first pair of candidate locations and the second candidate location of the second pair of candidate locations.

4. The method of claim 3 , wherein at least one of step i) and ii) comprises:
determining a compatibility score for the respective pair of candidate locations based upon the comparison of step (f), and
making the respective selection based upon that determination.

5. The method of claim 3 , further comprising performing step (g) for further pairs of the candidate locations so as to form further triplets of candidate locations.

6. The method of claim 5 , further comprising:
determining compatibility scores for the first and second pairs of the candidate locations of each triplet; and
selecting a subset of the triplets based on the determined compatibility scores.

7. The method of claim 6 , wherein the step of selecting a subset of the triplets comprises ranking the triplets according to their corresponding compatibility scores and selecting a subset of the triplets based on the ranking.

8. The method of claim 6 , wherein the selection of a subset of the triplets based on the determined compatibility scores is performed using graph searching.

9. The method of claim 6 , further comprising, for each triplet in the subset of triplets, computing an initial pose estimate for the imaged object based on the respective triplet of candidate locations.

10. The method of claim 9 , further comprising, for each triplet in the subset of triplets:
creating an estimated image of the imaged object based on the initial pose estimate of that triplet;
comparing the estimated image with the image representing the scene including the imaged object; and
based on the comparison, determining and scoring a refined pose estimate for the triplet.

11. The method of claim 10 , further comprising selecting one of the refined pose estimates based on the scores of the refined pose estimates.

12. The method of claim 1 , wherein step (b) comprises, for each of the plurality of image elements:
deriving a descriptor associated with the image element; and
matching that descriptor to one or more candidate locations in the atlas.

13. The method of claim 1 , wherein the image is a depth map and each image element has an intensity value that represents a depth, and further wherein step (d) comprises determining the first distance using image element intensity values.

14. An apparatus or system arranged to perform the method of preceding claim 1 .

15. A non-transitory computer-readable medium comprising machine-readable instructions arranged, upon execution by one or more processors, to cause the one or more processors to carry out the method of claim 1 .

Assignees

• Toshiba Kk

Inventors

• Zach Christopher

Classifications

• G06V10/761

  Proximity, similarity or dissimilarity measures · CPC title

• G06F18/22

  Matching criteria, e.g. proximity measures · CPC title

• G06T7/73 Primary

  using feature-based methods · CPC title

• G06T7/0042 Primary

  Physics · mapped topic

• G06K9/52

  Physics · mapped topic