A method and system for extracting gait contour map in surveillance video

Abstract

The invention discloses a method and system for extracting a gait contour map in a monitoring video, belonging to the field of computer vision, comprising: decoding and obtaining a sequence of video frames and then caching; performing pedestrian detection on the video frames to obtain a snapshot of a pedestrian target and a frame of a pedestrian target , and normalize the target frame to obtain a square snapshot and cache it; extract the pedestrian re-identification features according to the pedestrian target snapshot and cache them, and realize the pedestrian target in the current video frame and the pedestrian target in the previous frame through feature matching Matching, so as to realize the tracking of the same pedestrian target; if the trajectory of the same pedestrian target between consecutive n frames is successfully tracked, the square snapshot of the pedestrian target in consecutive n frames is obtained, and the human figure is segmented and then binarized , to obtain and cache the gait contour map of the pedestrian target in n consecutive frames. The invention can improve the extraction effect of the gait contour figure in the monitoring video, so as to meet the requirement of gait recognition.

Description

technical field [0001] The invention belongs to the field of computer vision, and more specifically relates to a method for extracting a gait contour map in a surveillance video. Background technique [0002] Today, with the rapid development of society and technology, various industries have increasing demands for safety and convenience in work and life. Video surveillance technology has emerged and continues to develop. In terms of "safe cities", "smart communities", and public transportation control, Both need to add a large number of monitoring cameras. Followed by massive video data. [0003] Target identification in video surveillance has applications in criminal investigation, access control and attendance, self-service, etc. The current mainstream method is to identify targets based on biometric features, including face recognition, fingerprint recognition, iris recognition, gait recognition, etc. Among them, gait recognition allows the computer to automatically id...

AI Technical Summary

Problems solved by technology

However, most of the existing methods are based on the existing gait data sets to realize the extraction of gait contour maps. In the existing gait data sets, the number of targets is single and the scene is simple. Background subtraction to obtain gait contour map

In general video processing, the background subtraction method is used to distinguish the background to extract the gait contour map of the moving target. However, in the actual video surveillance scene, there are situations such as illumination changes, small shooting angles, and picture jitters, which are prone to voids, "" "ghosting", large area noise, ground shadow foreground and other phenomena

Secondly, in the actual surveillance video, there are often multiple targets, and the existing gait contour map extraction methods cannot guarantee to obtain the gait contour map sequence of the same target

In general, the existing gait contour map extraction methods do not perform well in surveillance video application scenarios with multiple targets and complex backgrounds, and it is difficult to meet the requirements of gait recognition for gait contour maps.

Images