System and method for object re-identification

Patent metadata

Field	Value
Publication number	US-9852340-B2
Application number	US-201514869791-A
Country	US
Kind code	B2
Filing date	Sep 29, 2015
Priority date	Sep 30, 2014
Publication date	Dec 26, 2017
Grant date	Dec 26, 2017

Abstract

A method of identifying, with a camera, an object in an image of a scene, by determining the distinctiveness of each of a number of attributes of an object of interest, independent of the camera viewpoint, determining the detectability of each of the attributes based on the relative orientation of a candidate object in the image of the scene, determining a camera setting for viewing the candidate object based on the distinctiveness of an attribute, so as to increase the detectability of the attribute, and capturing an image of the candidate object with the camera setting to determine the confidence that the candidate object is the object of interest.

First claim

The invention claimed is:

1. A method of identifying, with a camera, an object in an image of a scene, the method comprising the steps of:
determining a distinctiveness of each of a plurality of attributes of an object of interest, independent of a camera viewpoint, the distinctiveness describing a uniqueness of the attribute;
determining a detectability of each of the plurality of attributes of a candidate object in the image of the scene based on a relative orientation of the candidate object, the detectability describing a degree of certainty with which the attribute can be detected in an image of the candidate object;
determining a camera setting for viewing the candidate object based on the determined distinctiveness of at least one attribute, so as to increase the detectability of the at least one attribute; and
capturing an image of the candidate object with the determined camera setting to determine a confidence that the candidate object is the object of interest.

2. A method according to claim 1 , wherein the camera setting for viewing the candidate object is dependent upon a confidence that the candidate object is the object of interest.

3. A method according to claim 1 , wherein the plurality of attributes are soft biometrics.

4. A method according to claim 3 , wherein the soft biometrics are textual or verbal descriptions of the object of interest.

5. A method according to claim 1 , wherein the step of determining the distinctiveness of an attribute comprises the step of constructing a tuple comprising a probability of an attribute label for the object of interest, and a frequency of the attribute label in a population of the candidate objects.

6. A method according to claim 1 , wherein the step of determining the detectability of an attribute comprises the steps of:
determining a prior probability that the candidate object is the object of interest;
determining viewing conditions under which the image of the candidate object was captured; and
testing an attribute classifier on a set of test images of different objects with the attribute captured under said viewing condition.

7. A method according to claim 1 , wherein the step of determining the camera setting for viewing the candidate object comprises the steps of:
selecting a provisional camera setting;
predicting detectability of each attribute of the candidate object using the provisional camera setting;
determining an increase in information about an identity of the candidate object observed using the provisional camera setting; and
determining the camera setting for viewing the candidate object dependent upon maximizing the increase in information.

8. A method according to claim 7 , wherein the increase in information is dependent upon mutual information between the observed attributes in the said camera setting and the identity of the candidate object, and the mutual information is weighted based on a confidence that the candidate object is the object of interest.

9. A method according to claim 1 , wherein the step of determining a camera setting for viewing the candidate object excludes any camera setting that leads to observing a previously observed attribute within a threshold period of time.

10. An apparatus comprising:
a camera for capturing an image of an object of interest and an image of a candidate object in a scene;
a processor; and
a memory storing a computer executable software program for directing the processor to perform a method for identifying, with the camera, an object in the image of the scene, the method comprising the steps of:
determining a distinctiveness of each of a plurality of attributes of the object of interest, independent of a camera viewpoint, the distinctiveness describing a uniqueness of the object;
determining a detectability of each of the plurality of attributes of a candidate object in the image of the scene based on a relative orientation of the candidate object, the detectability describing a degree of certainty with which the attribute can be detected in an image of the candidate object;
determining a camera setting for viewing the candidate object based on the determined distinctiveness of at least one attribute, so as to increase the detectability of the at least one attribute; and
capturing an image of the candidate object with the determined camera setting to determine a confidence that the candidate object is the object of interest.

11. A computer readable non-transitory memory storing a computer executable software program for directing processor to perform a method for identifying, with a camera, an object in an image of a scene, the method comprising the steps of:
determining a distinctiveness of each of a plurality of attributes of the object of interest, independent of a camera viewpoint, the distinctiveness describing a uniqueness of the attribute;
determining a detectability of each of the plurality of attributes of a candidate object in the image of the scene based on a relative orientation of the candidate object;
determining a camera setting for viewing the candidate object based on the determined distinctiveness of at least one attribute, so as to increase the detectability of the at least one attribute, the detectability describing a degree of certainty with which the attribute can be detected in an image of a candidate object; and
capturing an image of the candidate object with the determined camera setting to determine a confidence that the candidate object is the object of interest.

Assignees

• C/O Canon Kk
• Canon Kk

Inventors

• Mai fei
• Taylor Geoffrey Richard

Classifications

• H04N7/18

  Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast · CPC title

• G06V40/10 Primary

  Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands · CPC title

• G06F18/211

  Selection of the most significant subset of features · CPC title

• G06V10/24

  Aligning, centring, orientation detection or correction of the image · CPC title

• G06K9/00362

  Physics · mapped topic