# US 2015/0104066 A1 · Method for improving tracking in crowded situations using rival compensation

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2015/0104066 A1
- Document ID: US20150104066A1
- Status: Published application
- Kind: A1
- Application number: 14/508,955
- Filed: 2014-10-07
- Published: 2015-04-16
- Earliest claimed: 2013-10-10
- Assignees: Canon Kabushiki Kaisha
- Applicants: Canon Kabushiki Kaisha
- Inventors: Andrew John Shellshear
- Web page: https://patentgrep.com/patent/US20150104066A1

## CPC

- G06T 7/277: involving stochastic approaches, e.g. using Kalman filters
- G06V 20/53: {Recognition of crowd images, e.g. recognition of crowd congestion}
- G06T 2207/30196: Human being; Person
- G06V 10/50: by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

A method for tracking objects across a number of image frames includes tracking objects in the frames based on appearance models of each foreground region corresponding to each of the objects and determining if a plurality of the tracked objects overlap. Where a plurality of the tracked objects overlap, the method creates compensated appearance models for each of the plurality of overlapping objects by attenuating common appearance features among the corresponding appearance models; and tracks the plurality of overlapping objects based on the created compensated appearance models.

## Claims

### Claim 1

A method for tracking objects across a number of image frames, the method comprising:
tracking objects in the frames based on appearance models of each foreground region corresponding to each of the objects;
determining if a plurality of the tracked objects overlap;
where a plurality of the tracked objects overlap, creating compensated appearance models for each of the plurality of overlapping objects by attenuating common appearance features among the corresponding appearance models; and
tracking the plurality of overlapping objects based on the created compensated appearance models.

### Claim 2 (depends on 1)

A method according to claim 1, wherein:
the creating compensated appearance models further comprises, for a current tracked object being one of the overlapping objects, creating a rival compensated appearance model for the remaining overlapping objects; and
the tracking of the plurality of overlapping objects comprises, for the current tracked object, comparing the compensated appearance model of the current tracked object with the rival compensated appearance model of the remaining overlapping objects.

### Claim 3 (depends on 1)

A method according to claim 1, wherein:
the tracking of objects in the frames comprises, for a first object, obtaining and storing an exemplar view for the first object and an associated appearance model thereof in a first image frame; and
the creating compensated appearance models comprises forming, in association with a second image frame, an adjusted appearance model of the first object based on the associated appearance model of the first object and an appearance model of at least one nearby second object;
the method further comprising:
determining a scale and centroid adjustment of the exemplar view based on the adjusted appearance model;
applying tracking to the first object in the second image to determine a position of the first object in the second image; and
correcting the position of the first object in the second image according to the determined scale and centroid adjustment.

### Claim 4

A method for tracking a first object across a number of images, the method comprising:
obtaining and storing an exemplar view for the first object and an associated appearance model thereof in a first image;
forming, in association with a second image, an adjusted appearance model of the first object based on the associated appearance model of the first object and an appearance model of at least one nearby second object;
determining a scale and centroid adjustment of the exemplar view based on the adjusted appearance model;
applying tracking to the first object in the second image to determine a position of the first object in the second image; and
correcting the position of the first object in the second image according to the determined scale and centroid adjustment.

### Claim 5 (depends on 4)

A method according to claim 4, in which the adjusted appearance model comprises a compensated appearance model formed by attenuating common appearance features amongst the corresponding appearance models of at least the nearby second object.

### Claim 6

A method for tracking a first object across a number of images, the method comprising:
obtaining and storing an exemplar view for the first object and an associated appearance model thereof in a first image;
forming, in association with a second image, an adjusted appearance model of the first object based on the associated appearance model of the first object and an appearance model of at least one nearby second object;
determining a centroid adjustment of the exemplar view based on the adjusted appearance model;
applying tracking to the first object in the second image to determine a position of the first object in the second image; and
correcting the position of the first object in the second image according to the determined centroid adjustment.

### Claim 7 (depends on 6)

A method according to claim 6, in which the adjusted appearance model comprises a compensated appearance model formed by attenuating common appearance features amongst the corresponding appearance models of at least the nearby second object.

### Claim 8

A non-transitory computer readable storage medium having a program recorded thereon, the program being executable by a processor to track objects across a number of image frames, the program comprising:
code for tracking objects in the frames based on appearance models of each foreground region corresponding to each of the objects;
code for determining if a plurality of the tracked objects overlap;
code for creating, where a plurality of the tracked objects overlap, compensated appearance models for each of the plurality of overlapping objects by attenuating common appearance features among the corresponding appearance models; and
code for tracking the plurality of overlapping objects based on the created compensated appearance models.

### Claim 9 (depends on 8)

A non-transitory computer readable storage medium according to claim 8, wherein:
the code for creating further comprises, for a current tracked object being one of the overlapping objects, code for creating a rival compensated appearance model for the remaining overlapping objects; and
the code for tracking of the plurality of overlapping objects comprises, for the current tracked object, code for comparing the compensated appearance model of the current tracked object with the rival compensated appearance model of the remaining overlapping objects.

### Claim 10 (depends on 8)

A non-transitory computer readable storage medium according to claim 8, wherein:
the code for tracking of objects in the frames comprises code operative on a first object, for obtaining and storing an exemplar view for the first object and an associated appearance model thereof in a first image frame; and
the code for creating compensated appearance models comprises code for forming, in association with a second image frame, an adjusted appearance model of the first object based on the associated appearance model of the first object and an appearance model of at least one nearby second object;
the program further comprising:
code for determining a scale and centroid adjustment of the exemplar view based on the adjusted appearance model;
code for applying tracking to the first object in the second image to determine a position of the first object in the second image; and
code for correcting the position of the first object in the second image according to the determined scale and centroid adjustment.

### Claim 11

A non-transitory computer readable storage medium having a program recorded thereon, the program being executable by a processor to perform a method to track a first object across a number of images, the method comprising:
obtaining and storing an exemplar view for the first object and an associated appearance model thereof in a first image;
forming, in association with a second image, an adjusted appearance model of the first object based on the associated appearance model of the first object and an appearance model of at least one nearby second object;
determining a scale and centroid adjustment of the exemplar view based on the adjusted appearance model;
applying tracking to the first object in the second image to determine a position of the first object in the second image; and
correcting the position of the first object in the second image according to the determined scale and centroid adjustment.

### Claim 12 (depends on 11)

A non-transitory computer readable storage medium according to claim 11, in which the adjusted appearance model comprises a compensated appearance model formed by attenuating common appearance features amongst the corresponding appearance models of at least the nearby second object.

### Claim 13

A non-transitory computer readable storage medium having a program recorded thereon, the program being executable by a processor to perform a method to track a first object across a number of images, the method comprising:
obtaining and storing an exemplar view for the first object and an associated appearance model thereof in a first image;
forming, in association with a second image, an adjusted appearance model of the first object based on the associated appearance model of the first object and an appearance model of at least one nearby second object;
determining a centroid adjustment of the exemplar view based on the adjusted appearance model;
applying tracking to the first object in the second image to determine a position of the first object in the second image; and
correcting the position of the first object in the second image according to the determined centroid adjustment.

### Claim 14 (depends on 13)

A non-transitory computer readable storage medium according to claim 13, in which the adjusted appearance model comprises a compensated appearance model formed by attenuating common appearance features amongst the corresponding appearance models of at least the nearby second object.

### Claim 15

Computer apparatus comprising a processor and a memory upon which is recorded a program, the program being executable by the processor to track objects across a number of image frames, the program comprising:
code for tracking objects in the frames based on appearance models of each foreground region corresponding to each of the objects;
code for determining if a plurality of the tracked objects overlap;
code for creating, where a plurality of the tracked objects overlap, compensated appearance models for each of the plurality of overlapping objects by attenuating common appearance features among the corresponding appearance models; and
code for tracking the plurality of overlapping objects based on the created compensated appearance models.

### Claim 16

A video surveillance system comprising:
at least one camera configured to capture image frames of a scene;
computer apparatus coupled to the at least one camera, the computer apparatus comprising a processor and a memory upon which is recorded a program, the program being executable by the processor to track objects across a number of the image frames, captured by the at least one camera, the program comprising:
code for tracking objects in the frames based on appearance models of each foreground region corresponding to each of the objects;
code for determining if a plurality of the tracked objects overlap;
code for creating, where a plurality of the tracked objects overlap, compensated appearance models for each of the plurality of overlapping objects by attenuating common appearance features among the corresponding appearance models; and
code for tracking the plurality of overlapping objects based on the created compensated appearance models.
