Self-adaptive multi-target video tracking system under different scenes

Abstract

The invention provides a self-adaptive multi-target video tracking system under different scenes, aims to solve the problem of simultaneous tracking of a plurality of targets in video monitoring and manage the tracks of the tracking targets, and provides a multi-target tracking method based on single-target tracking improved fusion. A continuous adaptive mean shift algorithm and an online learning detection tracking algorithm are adopted, the two algorithms are applied to different monitoring scenes to be improved and fused to perform multi-target tracking respectively, and the multi-target video tracking system tracks and solves the problems of automatically extracting, detecting and tracking targets, and classifying the targets, different algorithms are used for tracking different targets, and meanwhile, post-processing and unified management are performed on the tracked trajectory result, so that the retrieval efficiency of the system on the target trajectory is greatly improved while the tracking effect is improved.

Description

technical field [0001] The invention relates to a multi-target video tracking system, in particular to an adaptive multi-target video tracking system in different scenarios, and belongs to the technical field of target video tracking. Background technique [0002] At the national strategic level, with the vigorous promotion of the construction of safe cities in various places, people's expectations and requirements for social security are getting higher and higher. With the expansion of the scope of video surveillance and the complexity of the environment, the wisdom of social comprehensive management business The video surveillance system has higher and higher requirements for target tracking. Numerous comprehensive management and security systems based on video surveillance have been deployed all over the country, and surveillance cameras have been installed in many important areas and road sections. However, the problem that followed at the beginning is that the relevant ...

AI Technical Summary

Problems solved by technology

[0013] Aiming at the deficiencies of the prior art, the present invention mainly solves the problems of tracking loss caused by object deformation and occlusion in the monitoring scene, the existing multi-target tracking algorithm cannot meet the real-time requirements of the tracking system, and the dynamic changes of tracking targets lead to track confusion and breakage, etc. The continuous self-adaptive mean shift algorithm of the present invention is based on color features, has fast tracking speed, and can adapt to multi-scale changes; the online learning detection tracking algorithm combines three processes of detection, learning and tracking to deal with the problems of occlusion and tracking loss in the monitoring environment The effect is good. The present invention improves and fuses these two algorithms and applies them to a multi-target video tracking system. According to different monitoring scene characteristics, an adaptive tracking algorithm is intelligently selected to achieve the best multi-target video tracking effect. The invention aims at multi-target video tracking features, and designs and implements a multi-target track processing unit. In the monitoring environment, the types of tracked targets are very complex and the number is constantly changing. The target moving tracks are classified and stored and managed in a unified manner. The multi-target trajectory processing unit can effectively solve the problem of tracking drift in the process of target tracking, and improve the accuracy of target tracking by post-processing the target trajectory

Images