Methods and apparatuses for searching for target person, devices, and media

1. A method for searching for a target person, comprising:

obtaining an image of the target person;

searching a face image library by using the image of the target person to obtain a first face image template matching the image of the target person, wherein the face image library comprises a plurality of face image templates; and

obtaining, according to the first face image template and a pedestrian image library, at least one target pedestrian image template matching the image of the target person, wherein the pedestrian image library comprises a plurality of pedestrian image templates.

2. The method according to claim 1 , wherein the searching a face image library by using the image of the target person to obtain a first face image template matching the image of the target person comprises:

obtaining first face feature data of the image of the target person; and

determining, according to the first face feature data, the first face image template from the plurality of face image templates stored in the face image library.

3. The method according to claim 2 , wherein the determining, according to the first face feature data, the first face image template from the plurality of face image templates stored in the face image library comprises:

determining the first face image template from the plurality of face image templates based on a distance between the first face feature data and each of a plurality of pieces of second face feature data corresponding to the plurality of face image templates.

4. The method according to claim 3 , wherein the determining the first face image template from the plurality of face image templates based on a distance between the first face feature data and each of a plurality of pieces of second face feature data corresponding to the plurality of face image templates comprises:

determining at least one face image template, which corresponds to second face feature data with a minimum distance from the first face feature data among the plurality of pieces of second face feature data, as the first face image template; or

determining at least one face image template, which corresponds to second face feature data with a distance less than or equal to a first threshold from the first face feature data among the plurality of pieces of second face feature data, as the first face image template.

5. The method according to claim 3 , wherein at least two of the plurality of face image templates form a face track corresponding to a same person, and second face feature data corresponding to the at least two face image templates is average feature data of the face track formed by the at least two face image templates.

6. The method according to claim 1 , wherein the obtaining, according to the first face image template and a pedestrian image library, at least one target pedestrian image template matching the image of the target person comprises:

obtaining a first pedestrian image corresponding to the first face image template; and

searching the pedestrian image library by using the first pedestrian image to obtain the at least one target pedestrian image template matching the image of the target person.

7. The method according to claim 6 , wherein the searching the pedestrian image library by using the first pedestrian image to obtain the at least one target pedestrian image template matching the image of the target person comprises:

determining first pedestrian feature data of the first pedestrian image;

determining, according to the first pedestrian feature data, at least one pedestrian image template matching the first pedestrian image from the plurality of pedestrian image templates stored in the pedestrian image library; and

determining the at least one pedestrian image template matching the first pedestrian image as the target pedestrian image template.

8. The method according to claim 7 , wherein the determining, according to the first pedestrian feature data, at least one pedestrian image template matching the first pedestrian image from the plurality of pedestrian image templates stored in the pedestrian image library comprises:

determining at least one pedestrian image template matching the first pedestrian image from the plurality of pedestrian image templates based on a distance between the first pedestrian feature data and each of a plurality of pieces of second pedestrian feature data corresponding to the plurality of pedestrian image templates.

9. The method according to claim 8 , wherein the determining at least one pedestrian image template matching the first pedestrian image from the plurality of pedestrian image templates based on a distance between the first pedestrian feature data and each of a plurality of pieces of second pedestrian feature data corresponding to the plurality of pedestrian image templates comprises:

determining at least one pedestrian image template, which corresponds to second pedestrian feature data with a minimum distance from the first pedestrian feature data among the plurality of pieces of second pedestrian feature data, as at least one pedestrian image template matching the first pedestrian image; or

determining at least one pedestrian image template, which corresponds to second pedestrian feature data with a distance less than or equal to a second threshold from the first pedestrian feature data among the plurality of pieces of second pedestrian feature data, as at least one pedestrian image template matching the first pedestrian image.

10. The method according to claim 8 , wherein at least two of the plurality of pedestrian image templates form a pedestrian track corresponding to a same person, and second pedestrian feature data corresponding to the at least two pedestrian image templates is average feature data of the pedestrian track formed by the at least two pedestrian image templates.

11. The method according to claim 6 , wherein the first face image template corresponds to a first face image; and

the obtaining a first pedestrian image corresponding to the first face image template comprises:

determining a first video to which the first face image belongs as well as frame number information and image position information of the first face image; and

obtaining a first pedestrian image corresponding to the first face image template from the first video according to the frame number information and the image position information of the first face image.

12. The method according to claim 11 , wherein the obtaining a first pedestrian image corresponding to the first face image template from the first video according to the frame number information and the image position information of the first face image comprises at least one of:

if the first video comprises a pedestrian image having a frame number which corresponds to the frame number information of the first face image and comprising the first face image, determining the pedestrian image comprising the first face image as a first pedestrian image corresponding to the first face image template; or

if the first video does not comprise a pedestrian image having a frame number which corresponds to the frame number information of the first face image and comprising the first face image, expanding the first face image in a first video image according to a preset proportion to obtain a first pedestrian image corresponding to the first face image template, wherein the frame number of the first video image in the first video corresponds to the frame number information of the first face image.

13. The method according to claim 1 , wherein after the obtaining, according to the first face image template and a pedestrian image library, at least one target pedestrian image template matching the image of the target person, the method further comprises:

filtering the at least one target pedestrian image template to obtain a search result of the target person.

14. The method according to 13 , wherein the filtering the at least one target pedestrian image template to obtain a search result of the target person comprises:

dividing the at least one target pedestrian image template into at least one group of pedestrian image templates according to feature data of each of the at least one target pedestrian image template;

determining filtering statistics data of each group of pedestrian image templates according to feature data of pedestrian images comprised in each of the at least one group of pedestrian image templates; and

determining a pedestrian image corresponding to one or more of the at least one group of pedestrian image templates as a search result of the target person according to the filtering statistics data of each of the at least one group of pedestrian image templates.

15. The method according to claim 1 , wherein before the searching a face image library by using the image of the target person to obtain a first face image template matching the image of the target person, the method further comprises:

performing face detection processing on video images comprised in each of at least one video to obtain a plurality of face images; and

storing each of the plurality of face images and/or information of each face image in the face image library as a face image template, wherein the information of the face image comprises at least one of the following: feature data of the face image, information of a video to which the face image belongs, frame number information of the face image, or position information of the face image in a video image.

16. The method according to claim 15 , further comprising:

performing face tracking processing on the plurality of face images to obtain at least one face track, where each face track comprises at least two of the plurality of face images;

determining average feature data of each face track of the at least one face track according to the at least two face images comprised in the each face track; and

storing the average feature data of each of the at least one face track in the face image library.

17. The method according to claim 1 , wherein before the obtaining, according to the first face image template and a pedestrian image library, at least one target pedestrian image template matching the image of the target person, the method further comprises:

performing dace detection processing on video images comprised in each of at least one video to obtain a plurality of pedestrian images; and

storing each of the plurality of pedestrian images and/or information of each pedestrian image in the pedestrian image library as a pedestrian image template, wherein the information of the pedestrian image comprises at least one of the following: feature data of the pedestrian image, information of a video to which the pedestrian image belongs, frame number information of the pedestrian image, and position information of the pedestrian image in the video image.

18. The method according to claim 17 , further comprising:

performing pedestrian tracking processing on the plurality of pedestrian images to obtain at least one pedestrian track, wherein each pedestrian track comprises at least two of the plurality of pedestrian images;

determining average feature data of each pedestrian track of the at least one pedestrian track according to the at least two pedestrian images comprised in the each pedestrian track; and

storing the average feature data of each of the at least one pedestrian track in the pedestrian image library.

19. An apparatus for searching for a target person, comprising:

a processor; and

a memory for storing instructions executable by the processor,

wherein execution of the instructions by the processor causes the processor to implement operations, the operations comprising:

obtaining an image of the target person;

searching a face image library by using the image of the target person to obtain a first face image template matching the image of the target person, wherein the face image library comprises a plurality of face image templates; and

obtaining, according to the first face image template and a pedestrian image library, at least one target pedestrian image template matching the image of the target person, wherein the pedestrian image library comprises a plurality of pedestrian image templates.

20. A non-transitory computer storage medium, having computer-readable instructions stored thereon, wherein execution of the computer-readable instructions by a processor causes the processor to implement operation of:

obtaining an image of the target person;

searching a face image library by using the image of the target person to obtain a first face image template matching the image of the target person, wherein the face image library comprises a plurality of face image templates; and

obtaining, according to the first face image template and a pedestrian image library, at least one target pedestrian image template matching the image of the target person, wherein the pedestrian image library comprises a plurality of pedestrian image templates.