METHOD AND SYSTEM FOR PARTIAL OCCLUSION HANDLING IN VEHICLE TRACKING USING DEFORMABLE PARTS MODEL

• US 20150310624A1
• Filed: 02/26/2015
• Published: 10/29/2015
• Est. Priority Date: 04/24/2014
• Status: Active Grant

• Current Assignee: Conduent Business Services, LLC (Conduent, Inc.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Bulan, Orhan, Burry, Aaron Michael, Artan, Yusuf Oguzhan
• Granted Patent: US 9,779,331 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06T 2207/30232 Surveillance
• CPC Class Codes: G06T 2207/30236 Traffic on road, railway or crossing
• CPC Class Codes: G06T 7/215 Motion-based segmentation
• CPC Class Codes: G06T 7/248 involving reference images or patches
• CPC Class Codes: G06T 7/254 involving subtraction of images
• CPC Class Codes: G06V 20/54 of traffic, e.g. cars on the road, trains or boats

Abstract

Provided is a method and system of tracking partially occluded objects using an elastic deformation model. According to an exemplary method and system, partially occluded vehicles are detected and tracked in a scene including side-by-side drive-thru lanes. A method for updating an event sequence includes acquiring video data of a queue area from at least one image source; searching the frames for subjects located at least near a region of interest (ROI) of defined start points in the video data; tracking a movement of each detected subject through the queue area over a subsequent series of frames; using the tracking, determining if a location of the a tracked subject reaches a predefined merge point where multiple queues in the queue area converge into a single queue lane; in response to the tracked subject reaching the predefined merge point, computing an observed sequence of where the tracked subject places among other subjects approaching an end-event point; and, updating a sequence of end-events to match the observed sequence of subjects in the single queue lane.

Claims

1. A computer implemented method of tracking a partially occluded object in a scene captured with an image capturing device, the method comprising:

a) continuously monitoring an area of interest via an image capturing device;

b) identifying candidate frames captured with the image capturing device, one or more of the candidate frames including a vehicle within one or more defined ROIs (Region of Interest);

c) detecting a non-occluded portion of the vehicle in one or more of the candidate frames using one or more classifiers trained using extracted features from non-occluded portions of a set of training samples; and

d) tracking the non-occluded portion of the vehicle across a plurality of video frames.

2. The computer implemented method according to claim 1, wherein prior to step d), the method comprises:

checking if the detected non-occluded portion of the vehicle is already being tracked.

3. The computer implemented method according to claim 1, wherein prior to step b), the method comprises:

defining region of interest (ROI) associated with a field of view (FOV) associated with the image capturing device.

4. The computer implemented method according to claim 3, wherein prior to step b), the method comprises:

defining two or more ROIs for automated sequencing of vehicles in a side-by-side drive-thru configuration.

5. The computer implemented method according to claim 1, wherein the candidate frames are identified using a background subtraction process including a Gaussian mixture model, eigenbackgrounds, principle component analysis or running average computation.

6. The computer implemented method according to claim 1, wherein the candidate frames are identified using a motion detection process with the ROI including a temporal difference process, a pixel-level optical flow process or a block-matching process.

7. The computer implemented method according to claim 1, wherein step c) includes using one or more deformable parts models and the one or more deformable parts models are trained using positive video samples including a partially occluded vehicle and a plurality of landmarks are assigned to the positive video sample which outline a non-occluded portion of the vehicle.

8. An image processing system comprising:

a memory and a processor operatively associated with the memory configured to receive a digital image of a scene and execute instructions to perform a method of tracking a partially occluded object in a scene captured with an image capturing device, the method comprising;

a) continuously monitoring an area of interest via an image capturing device;

b) identifying candidate frames captured with the image capturing device, one or more of the candidate frames potentially including a vehicle within one or more defined ROIs (Region of Interest);

c) detecting a non-occluded portion of the vehicle in one or more of the candidate frames using one or more classifiers trained using extracted features from non-occluded portions of a set of training samples; and

d) tracking the non-occluded portion of the vehicle across a plurality of video frames.

9. The image processing system according to claim 8, wherein prior to step d), the method comprises:

checking if the detected non-occluded portion of the vehicle is already being tracked.

10. The image processing system according to claim 8, wherein prior to step b), the method comprises:

defining region of interest (ROI) associated with a field of view (FOV) associated with the image capturing device.

11. The image processing system according to claim 10, wherein prior to step b), the method comprises:

defining two or more ROIs for automated sequencing of vehicles in a side-by-side drive-thru configuration.

12. The image processing system according to claim 8, wherein the candidate frames are identified using a background subtraction process including a Gaussian mixture model, eigenbackgrounds, principle component analysis or running average computation.

13. The image processing system according to claim 8, wherein the candidate frames are identified using a motion detection process with the ROI including a temporal difference process, a pixel-level optical flow process or a block-matching process.

14. The image processing system according to claim 8, wherein step c) includes using one or more deformable parts models and the one or more deformable parts models are trained using positive video samples including a partially occluded vehicle and a plurality of landmarks are assigned to the positive video sample which outline a non-occluded portion of the vehicle.

15. A system for updating an event sequence, the system comprising an automatic sequencing device including a memory and a processor in communication with the processor configured to:

acquire video data of a queue area from at least one image source, each image source associated with an image plane;

receive a number of individual event requests from multiple subjects, each event request being received when a subject is located in one of multiple queues in the queue area;

process the video data using an elastic deformation model to detect a subject located near at least one of a plurality of start points, each start point corresponding to a location associated with one of the multiple queues in the video data;

track a movement of each detected subject through the queue area over a subsequent series of frames;

determine if a location of a tracked subject reaches a predefined merge point on the image plane where the multiple queues in the queue area converge into a single queue lane;

in response to the tracked subject reaching the predefined merge point, compute an observed sequence of where the tracked subject places relative to other subjects already past the merge point and approaching an end-event point in the single queue lane; and

,

update a sequence of the events to match the observed sequence of subjects in the single queue lane.

16. The system for updating an event sequence according to claim 15, wherein the detected and tracked subject is partially occluded by another subject, and a non-occluded portion of the subject is detected and tracked by the elastic deformation model.

in response to detecting an object-in-motion, associate the detected object-in-motion as a candidate subject.

17. The system of updating an event sequence according to claim 16, wherein the memory and processor is further configured to:

process the acquired video data to determine candidate subjects;

classify each candidate subject as belonging to one of a new subject and a currently tracked subject;

in response to the candidate subject being a new subject, assign a tracker to the new subject; and

in response to the candidate subject being a currently tracked subject, discard the candidate subject.

18. The system for updating an event sequence according to claim 17, wherein the candidate subjects are determined from candidate video frames which are identified using one of a background subtraction process and a motion detection process.

19. The system of updating an event sequence according to claim 17, wherein the memory and processor is further configured to:

extract a feature descriptor from each detected candidate subject;

compare the extracted feature descriptor to one of a threshold and corresponding descriptors of current trackers; and

based on the comparing, classify each detected candidate subject as belonging to one of a new subject and a currently tracked subject.

20. The system of updating an event sequence according to claim 15, wherein the memory and processor is further configured to:

define an ROI in the queue area.

21. The system of updating an event sequence according to claim 15, wherein the queue area includes a fast-food drive-thru, the event-requests include orders, and the end-events include one of goods and services.

22. The system of updating an event sequence according to claim 15, wherein the processor is further configured to:

compute a statistic of tracked feature locations of the subject;

compare the statistic to a location of a merge point region; and

,

determine if a location of the tracked subject reaches the predefined merge point based on the comparison;

wherein the statistic is selected from a group consisting of;

a single point identifying a position of the tracked subject, a centroid of the tracked feature locations, a median center location of the tracked feature locations, a first point, a predetermined percentage of feature points, a last point, and a degree of overlap.