Granted Patent B2

US 12,283,106 · App. 17/304,740 · Granted Apr 22, 2025

Inventor: Yuntao Li (Hangzhou, CN)

Assignee: ZHEJIANG DAHUA TECHNOLOGY CO., LTD.

G06V20/52G06T7/248G06T7/74G06V20/40G06V40/161G06T2207/10016G06T2207/30201

Abstract

Claims (73)

1. A system for tracking an object in video surveillance, comprising:

at least one storage medium including a set of instructions for tracking the object in video surveillance; and

at least one processor in communication with the storage medium, wherein when executing the set of instructions, the at least one processor is directed to:

obtain a video including a plurality of consecutive frames;

obtain a current frame from the plurality of consecutive frames, wherein an object of interest is identified in at least two previous frames of the current frame;

determine a searching region in the current frame based on a reference frame associated with the current frame, wherein the searching region is an enlarged face recognition box in the current frame with respect to a face recognition box corresponding the object of interest of the reference frame, and the enlarged face recognition box is determined based on an enlargement degree;

obtain at least two template frames from the at least two previous frames, wherein the at least two template frames include a plurality of features of the object of interest; and

for each of the at least two template frames, determine at candidate face regions based on the searching region and the at least two template frames using a correlation filter to obtain at least two candidate face regions;

identify a position related to a face of the object of interest in the current frame based on the at least two candidate face regions; and

update the face of the object of interest in a later frame based on the at least two candidate face regions.

2. The system of claim 1 , wherein the at least one processor is further directed to:

identify the object of interest in one or more later frames of the current frame based on the identified position related to the object of interest in the current frame.

3. The system of claim 1 , wherein the object of interest is a person of interest, and the at least one processor is further directed to:

obtain one or more tracked faces from the video;

determine one or more face regions based on the current frame;

select a tracked face from the one or more tracked faces;

select a face region from the one or more face regions; and

determine whether the face region correlates to the tracked face.

4. The system of claim 3 , wherein the at least one processor is further directed to:

in response to a determination that the face region correlates to the tracked face, correlate the face region with a person of interest corresponding to the tracked face; and

update the tracked face in the current frame with the face region.

5. The system of claim 3 , wherein the at least one processor is further directed to:

in response to a determination that the face region does not correlate to the tracked face, identify the position related to a person of interest in the current frame based on the correlation filter; and

update the tracked face in the current frame with the identified position.

6. The system of claim 3 , wherein to determine whether the face region correlates to the tracked face, the at least one processor is further directed to:

determine whether a percentage of overlapping area between the face region and the tracked face exceeds an overlapping threshold; and

in response to a determination that the percentage of overlapping area exceeds the overlapping threshold, determine that the face region correlates to the tracked face.

7. The system of claim 1 , wherein the object of interest is a person of interest, and to obtain the at least two template frames, the at least one processor is further directed to:

determine a matching period associated with the person of interest based on previous frames of the current frame; and

determine the at least two template frames based on the matching period, wherein the person of interest is identified in the at least two template frames.

8. The system of claim 1 , wherein the reference frame is a previous frame next to the current frame.

9. The system of claim 7 , wherein the at least two template frames include a first frame in the matching period, a middle frame in the matching period, and a last frame in the matching period.

10. The system of claim 1 , wherein the determining the searching region includes:

determining the enlargement degree; and

determining the searching region based on the reference frame and the enlargement degree.

11. The system of claim 1 , wherein the enlargement degree is adjustable.

12. A method for tracking an object in video surveillance, comprising:

obtaining a video including a plurality of consecutive frames;

obtaining a current frame from the plurality of consecutive frames, wherein an object of interest is identified in at least two previous frames of the current frame;

determining a searching region in the current frame based on a reference frame associated with the current frame, wherein the searching region is an enlarged face recognition box in the current frame with respect to a face recognition box corresponding the object of interest of the reference frame, and the enlarged face recognition box is determined based on an enlargement degree;

obtaining at least two template frames from the at least two previous frames, wherein the at least two template frames include a plurality of features of the object of interest; and

for each of the at least two template frames, determine at candidate face regions based on the searching region and the at least two template frames using a correlation filter to obtain at least two candidate face regions;

identifying a position related to a face of the object of interest in the current frame based on the at least two candidate face regions; and

updating the face of the object of interest in a later frame based on the at least two candidate face regions.

13. The method of claim 12 further comprising:

identifying the object of interest in one or more later frames of the current frame based on the identified position related to the object of interest in the current frame.

14. The method of claim 12 , wherein the object of interest is a person of interest, and the method further includes:

obtaining one or more tracked faces from the video;

determining one or more face regions based on the current frame;

selecting a tracked face from the one or more tracked faces;

selecting a face region from the one or more face regions; and

determining whether the face region correlates to the tracked face.

15. The method of claim 14 further comprising:

in response to a determination that the face region correlates to the tracked face, correlating the face region with a person of interest corresponding to the tracked face; and

updating the tracked face in the current frame with the face region.

16. The method of claim 14 further comprising:

in response to a determination that the face region does not correlate to the tracked face, identifying the position related to a person of interest in the current frame based on the correlation filter; and

updating the tracked face in the current frame with the identified position.

17. The method of claim 14 , wherein the determining whether the face region correlates to the tracked face includes:

determining whether a percentage of overlapping area between the face region and the tracked face exceeds an overlapping threshold; and

in response to a determination that the percentage of overlapping area exceeds the overlapping threshold, determining that the face region correlates to the tracked face.

18. The method of claim 12 , wherein the object of interest is a person of interest, and the obtaining the at least two template frames includes:

determining a matching period associated with the person of interest based on previous frames of the current frame; and

determining the at least two template frames based on the matching period, wherein the person of interest is identified in the at least two template frames.

19. The method of claim 18 , wherein the at least two template frames include a first frame in the matching period, a middle frame in the matching period, and a last frame in the matching period.

20. A non-transitory computer readable medium, comprising at least one set of instructions for tracking an object in video surveillance, wherein when executed by at least one processor of one or more electronic device, the at least one set of instructions directs the at least one processor to:

obtain a video including a plurality of consecutive frames;

obtain a current frame from the plurality of consecutive frames, wherein an object of interest is identified in at least two previous frames of the current frame;

determine a searching region in the current frame based on a reference frame associated with the current frame, wherein the searching region is an enlarged face recognition box in the current frame with respect to a face recognition box corresponding the object of interest of the reference frame, and the enlarged face recognition box is determined based on an enlargement degree;

obtain at least two template frames from the at least two previous frames, wherein the at least two template frames include a plurality of features of the object of interest; and

for each of the at least two template frames, determine at candidate face regions based on the searching region and the at least two template frames using a correlation filter to obtain at least two candidate face regions;

identify a position related to a face of the object of interest in the current frame based on the at least two candidate face regions; and

update the face of the object of interest in a later frame based on the at least two candidate face regions.

Assignments (1)

Continuity (2)

References Cited (29)