AUTOMATIC TIME SIGNATURE-BASED VIDEO MATCHING FOR A CAMERA NETWORK

• US 20160014305A1
• Filed: 07/07/2015
• Published: 01/14/2016
• Est. Priority Date: 07/11/2014
• Status: Active Grant

• Current Assignee: AGT International GMBH
• Original Assignee: AGT International GMBH
• Inventors: SCHLATTMANN, Markus, Diener, Stefan
• Granted Patent: US 9,942,450 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06V 20/48 Matching video sequences
• CPC Class Codes: G06V 20/54 of traffic, e.g. cars on the road, trains or boats
• CPC Class Codes: H04N 5/04 Synchronising for television systems using pulse code modulation H04N7/56
• CPC Class Codes: H04N 5/44 Receiver circuitry for the reception of television signals according to analogue transmission standards H04N5/14 takes precedence

Abstract

A method for automatically matching video streams from two cameras of a camera network includes obtaining a video stream of frames that are acquired by each of the cameras. Each video stream includes images of moving objects. A time signature for each of the video streams is calculated. Each time signature is indicative of a time at which an image of one the objects is located at a predetermined part of the frame. A temporal offset of one of the signatures relative to the other signature is calculated such that, when applied to one of the signatures, a correspondence between the signatures is maximized. The temporal offset is applicable to video streams that are acquired by the two cameras to determine if a moving object that is imaged by one of the cameras is identical to a moving object that is imaged by the other camera.

Claims

1. A method for automatically matching video streams from two cameras of a camera network, the method comprising:

obtaining a video stream of frames that are acquired by each of the cameras, each video stream including images of a plurality of moving objects;

calculating a time signature for each of the video streams, each time signature being indicative of a time at which an image of one of said plurality of moving objects is located at a predetermined part of the frame; and

calculating a temporal offset of one of the signatures relative to the other signature that, when applied to said one of the signatures, substantially maximizes a correspondence between said one of the signatures and the other signature, the temporal offset being applicable to video streams that are acquired by the two cameras to determine if a moving object that is imaged by one of the cameras is identical to a moving object that is imaged by the other of the cameras.

2. The method of claim 1, wherein said plurality of moving objects moves along a substantially common path.

3. The method of claim 2, wherein the predetermined part is located on an image of the path.

4. The method of claim 2, wherein the path comprises a traffic lane.

5. The method of claim 2, wherein said plurality of moving objects moves along the path with a substantially constant speed.

6. The method of claim 1, further comprising changing a location of the predetermined part of the frame in accordance with a comparison of the time signatures.

7. The method of claim 1, wherein the predetermined part is defined by a point.

8. The method of claim 1, wherein calculating the time signature comprises applying a binary scheme.

9. The method of claim 1, wherein the correspondence is maximized when a correlation of the two signatures is maximized.

10. The method of claim 1, wherein calculating the temporal offset comprises applying dynamic time warping to the signatures.

11. The method of claim 1, further comprising using the temporal offset to calculate a spatial relationship between fields of view of the two cameras.

12. The method of claim 11, wherein calculating the spatial relationship comprises calculating a degree of overlap between the fields of view.

13. The method of claim 11, wherein calculating the spatial relationship comprises calculating a gap between the fields of view.

14. The method of claim 1, further comprising using the calculated temporal offset to calculate a de-synchronization of the video streams.

15. A non-transitory computer readable storage medium having stored thereon instructions that when executed by a processor will cause the processor to perform the method of:

obtaining a video stream that is acquired by each camera of a camera network of a plurality of moving objects;

calculating a time signature for each video stream, the time signature being indicative of a time at which an image of one of said plurality of moving objects is located at a predetermined part of a frame of that video stream; and

calculating a temporal offset between two of the time signatures, such that when the temporal offset is applied to one of the two time signatures, a correspondence between the two time signatures is maximized; and

applying the temporal offset to video streams that are acquired by the cameras that acquired the video streams from which said two of the time signatures were calculated to determine if a moving object that is imaged by one of the cameras is identical to a moving object that is imaged by the other of the cameras.

16. The computer readable storage medium of claim 15, wherein the instructions are further executable by the processor to identify within each video stream an image of a path along which said plurality of moving objects moves.

17. The computer readable storage medium of claim 15, wherein the instructions are further executable by the processor to cause the processor to perform the method of calculating a spatial relationship between fields of view of the cameras.

18. A system comprising;

a network of a plurality of cameras, each camera being configured to acquire a video stream of a plurality of moving objects; and

a processing unit in communication with a computer readable medium, wherein the computer readable medium contains a set of instructions wherein the processing unit is designed to carry out the set of instructions to;

calculate a time signature for each video stream that is acquired by each of said plurality of cameras, each time signature being indicative of a time at which an image of one of said plurality of moving objects in that video stream is located at a predetermined part of a frame of that video stream;

calculate a temporal offset each of the time signatures relative to another of the time signatures such that when the time offset is applied to said each of the time signatures, a correspondence between said each of the time signatures and said another of the time sinatures is maximized; and

apply the temporal offset to video streams that are acquired by different cameras of said plurality of cameras to determine if a moving object that is imaged by one of said plurality of cameras is identical to a moving object that is imaged by another of said plurality of cameras.

19. The system of claim 18, wherein said plurality of moving objects moves along a substantially common path.

20. The system of claim 18, wherein said plurality of moving objects comprises a plurality of vehicles.