Face detection and search method for a monitoring video

Abstract

The invention provides a monitoring video-oriented face detection and search method. The method comprises the following steps: firstly, training a face detector; Detecting an input monitoring video frame to be subjected to face recognition and search by using a face detector to obtain a face region If in the monitoring video frame, and positioning five sense organs in the face region If to obtaina monitoring video face five sense organ positioning result; Determining a target face image, and performing five-sense-organ positioning on the target face image to obtain a target face five-sense-organ positioning result; And then calculating the similarity between the whole face and the local five sense organs according to the five sense organ positioning result of the monitoring video face image and the five sense organ positioning result of the target face image obtained in the previous step. And finally, calculating probability fusion similarity of the monitoring video face image and thetarget face image to obtain a search matching result. By using the method, the search result is more accurate.

Description

technical field [0001] The invention relates to the technical field of face recognition, in particular to a video surveillance-oriented face detection and search method. Background technique [0002] With the large-scale application of video surveillance systems and in-depth research on related recognition algorithms, the combination of face recognition search based on face recognition technology and video surveillance can effectively improve the intelligence level of video surveillance, especially in network pursuit, Combating human trafficking and other aspects can improve work efficiency, reduce manual workload, and increase the probability of finding the target person. At present, a prominent restrictive factor of this application is: since the surveillance video mainly shoots a large scene, it is easy to cause the face area in it to be small, so that the face cannot be detected, resulting in poor recognition and search results. [0003] Existing face detection and sear...

AI Technical Summary

Problems solved by technology

[0004] The main work of the above methods is concentrated on the facial feature representation part, which cannot solve the problem of missing face detection caused by the small face area in the surveillance video, and the above method mainly extracts features for faces with large size and rich details, while For faces with small areas and low details in the actual surveillance video, the above feature extraction methods cannot obtain the effect commensurate with high-quality face images

Images