MULTI-TARGET PEDESTRIAN TRACKING METHOD, MULTI-TARGET PEDESTRIAN TRACKING APPARATUS AND MULTI-TARGET PEDESTRIAN TRACKING DEVICE

20220004747

17/280821

July 28, 2020

A multi-target pedestrian tracking method, a multi-target pedestrian tracking apparatus and a multi-target pedestrian tracking device are provided, related to the field of image processing technologies. The multi-target pedestrian tracking method includes: detecting a plurality of candidate pedestrian detection boxes in a current frame of image to be detected, where a temporary tracking identification and a tracking counter are set for each of the plurality of candidate pedestrian detection boxes; and determining whether each of the plurality of candidate pedestrian detection boxes matches an existing tracking box, updating a value of the tracking counter according to a determination result, and continuing to detect a next frame of image to be detected. When the value of the tracking counter reaches a first preset threshold, the updating the value of the tracking counter is stopped, and the temporary tracking identification is converted to a confirmed tracking identification.

BOE TECHNOLOGY GROUP CO., LTD. (Beijing, CN)

1. A multi-target pedestrian tracking method, comprising: detecting a plurality of candidate pedestrian detection boxes in a current frame of image to be detected, wherein a temporary tracking identification and a tracking counter are set for each of the plurality of candidate pedestrian detection boxes; and determining whether each of the plurality of candidate pedestrian detection boxes matches an existing tracking box, updating a value of the tracking counter according to a determination result, and continuing to detect a next frame of image to be detected, wherein in a case that the value of the tracking counter reaches a first preset threshold, the updating the value of the tracking counter is stopped, and the temporary tracking identification is converted to a confirmed tracking identification.

2. The multi-target pedestrian tracking method according to claim 1, wherein the updating the value of the tracking counter according to the determination result comprises: in a case that the candidate pedestrian detection box matches the existing tracking box, adding 1 to the value of the tracking counter; wherein the tracking counter is established after the candidate pedestrian detection box is detected for a first time, and an initial value of the tracking counter is 0.

3. The multi-target pedestrian tracking method according to claim 1, wherein the updating the value of the tracking counter according to the determination result comprises: in a case that the candidate pedestrian detection box matches the existing tracking box, adding 1 to the value of the tracking counter; in a case that the candidate pedestrian detection box does not match the existing tracking box, subtracting 1 from the value of the tracking counter; wherein the tracking counter is established after the candidate pedestrian detection box is detected for a first time, and an initial value of the tracking counter is an integer lager than 0.

4. The multi-target pedestrian tracking method according to claim 3, wherein in a case that the value of the tracking counter is smaller than a second preset threshold, deleting the candidate pedestrian detection box, wherein the second preset threshold is smaller than the first preset threshold.

5. The pedestrian tracking method according to claim 1, wherein the determining whether each of the plurality of candidate pedestrian detection boxes matches the existing tracking box comprises: calculating a feature distance between the candidate pedestrian detection box and tracking boxes in previous N frames of images of the current frame of image to be detected, in a case that the feature distance is smaller than a third preset threshold, determining that the candidate pedestrian detection box matches the existing tracking box, and in a case that the feature distance is larger than or equal to the third preset threshold, determining that the candidate pedestrian detection box does not match the existing tracking box, wherein N is an integer larger than 1.

6. The multi-target pedestrian tracking method according to claim 5, wherein the calculating the feature distance between the candidate pedestrian detection box and the tracking boxes in the previous N frames of images of the current frame of image to be detected comprises: calculating a feature of the candidate pedestrian detection box; calculating a distance dist(n) between the feature of the candidate pedestrian detection box and a feature of a tracking box in a previous nth frame of image of the current frame of image to be detected, wherein n is an integer larger than or equal to 1 and smaller than or equal to N; and calculating the feature distance Dmean through the following formula: [mathematical expression included]

7. The multi-target pedestrian tracking method according to claim 6, wherein subsequent to the detecting the plurality of candidate pedestrian detection boxes in the current frame of image to be detected, the method further comprises: for each of the plurality of candidate pedestrian detection boxes in the current frame of image to be detected, calculating partial intersection-over-unions between the candidate pedestrian detection box and other respective candidate pedestrian detection boxes; in a case that any one of partial intersection-over-unions is larger than a fourth preset threshold, not storing the feature of the candidate pedestrian detection box; and in a case that all partial intersection-over-unions are smaller than or equal to the fourth preset threshold, storing the feature of the candidate pedestrian detection box as a feature of a tracking box in a current frame of image.

8. The multi-target pedestrian tracking method according to claim 7, wherein the partial intersection-over-union is [mathematical expression included] wherein A is the candidate pedestrian detection box, and B is any one of other candidate pedestrian detection boxes.

9. A multi-target pedestrian tracking apparatus, comprising: a memory, a processor, and a computer program stored in the memory and executable by the processor, the processor is configured to execute the computer program to: detect a plurality of candidate pedestrian detection boxes in a current frame of image to be detected, wherein a temporary tracking identification and a tracking counter are set for each of the plurality of candidate pedestrian detection boxes; determine whether each of the plurality of candidate pedestrian detection boxes matches an existing tracking box, update a value of the tracking counter according to a determination result, and continue to detect a next frame of image to be detected; and in a case that the value of the tracking counter reaches a first preset threshold, stop updating the value of the tracking counter, and convert the temporary tracking identification to a confirmed tracking identification.

10. The multi-target pedestrian tracking apparatus according to claim 9, wherein the processor is configured to execute the computer program to: in a case that the candidate pedestrian detection box matches the existing tracking box, add 1 to the value of the tracking counter; wherein the tracking counter is established after the candidate pedestrian detection box is detected for a first time, and an initial value of the tracking counter is 0.

11. The multi-target pedestrian tracking apparatus according to claim 9, wherein the processor is configured to execute the computer program to: in a case that the candidate pedestrian detection box matches the existing tracking box, add 1 to the value of the tracking counter; in a case that the candidate pedestrian detection box does not match the existing tracking box, subtract 1 from the value of the tracking counter; wherein the tracking counter is established after the candidate pedestrian detection box is detected for a first time, and an initial value of the tracking counter is an integer lager than 0.

12. The multi-target pedestrian tracking apparatus according to claim 11, wherein the processor is configured to execute the computer program to: in a case that the value of the tracking counter is smaller than a second preset threshold, delete the candidate pedestrian detection box, wherein the second preset threshold is smaller than the first preset threshold.

13. The multi-target pedestrian tracking apparatus according to claim 9, wherein the processor is configured to execute the computer program to: calculate a feature distance between the candidate pedestrian detection box and tracking boxes in previous N frames of images of the current frame of image to be detected, in a case that the feature distance is smaller than a third preset threshold, determine that the candidate pedestrian detection box matches the existing tracking box, and in a case that the feature distance is larger than or equal to the third preset threshold, determine that the candidate pedestrian detection box does not match the existing tracking box, wherein N is an integer larger than 1.

14. The multi-target pedestrian tracking apparatus according to claim 13, wherein the processor is configured to execute the computer program to: calculate a feature of the candidate pedestrian detection box; calculate a distance dist(n) between the feature of the candidate pedestrian detection box and a feature of a tracking box in a previous nth frame of image of the current frame of image to be detected, wherein n is an integer larger than or equal to 1 and smaller than or equal to N; and calculate the feature distance Dmean through the following formula: [mathematical expression included]

15. The multi-target pedestrian tracking apparatus according to claim 14, wherein the processor is configured to execute the computer program to: for each of the plurality of candidate pedestrian detection boxes in the current frame of image to be detected, calculate partial intersection-over-unions between the candidate pedestrian detection box and other respective candidate pedestrian detection boxes, in a case that any one of partial intersection-over-unions is larger than a fourth preset threshold, not store the feature of the candidate pedestrian detection box, and in a case that all partial intersection-over-unions are smaller than or equal to the fourth preset threshold, store the feature of the candidate pedestrian detection box as a feature of a tracking box in a current frame of image.

16. The multi-target pedestrian tracking apparatus according to claim 15, wherein the partial intersection-over-union is [mathematical expression included] wherein A is the candidate pedestrian detection box, and B is any one of other candidate pedestrian detection boxes.

17. (canceled)

18. A computer-readable storage medium storing a computer program, wherein the computer program is executed by a processor to: detect a plurality of candidate pedestrian detection boxes in a current frame of image to be detected, wherein a temporary tracking identification and a tracking counter are set for each of the plurality of candidate pedestrian detection boxes; and determine whether each of the plurality of candidate pedestrian detection boxes matches an existing tracking box, update a value of the tracking counter according to a determination result, and continue to detect a next frame of image to be detected, wherein in a case that the value of the tracking counter reaches a first preset threshold, the updating the value of the tracking counter is stopped, and the temporary tracking identification is converted to a confirmed tracking identification.

19. The computer-readable storage medium according to claim 18, wherein the computer program is executed by the processor to: in a case that the candidate pedestrian detection box matches the existing tracking box, add 1 to the value of the tracking counter; wherein the tracking counter is established after the candidate pedestrian detection box is detected for a first time, and an initial value of the tracking counter is 0.

20. The computer-readable storage medium according to claim 18, wherein the computer program is executed by the processor to: in a case that the candidate pedestrian detection box matches the existing tracking box, add 1 to the value of the tracking counter; in a case that the candidate pedestrian detection box does not match the existing tracking box, subtract 1 from the value of the tracking counter; wherein the tracking counter is established after the candidate pedestrian detection box is detected for a first time, and an initial value of the tracking counter is an integer lager than 0.

21. The computer-readable storage medium according to claim 18, wherein the computer program is executed by the processor to: calculate a feature distance between the candidate pedestrian detection box and tracking boxes in previous N frames of images of the current frame of image to be detected, in a case that the feature distance is smaller than a third preset threshold, determine that the candidate pedestrian detection box matches the existing tracking box, and in a case that the feature distance is larger than or equal to the third preset threshold, determine that the candidate pedestrian detection box does not match the existing tracking box, wherein N is an integer larger than 1.

06; 06; 06

edspap.20220004747