Suspicious Personnel Tracking And Recognition System Based On Machine Vision

In the real-time monitoring process,traditional video surveillance systems have problems such as high labor cost,high missed detection rate,and untimely alarm.It is necessary to study key technical problems and systems in suspicious personnel tracking and identification.In this paper,the methods of target detection target tracking and behavior recognition in video analysis technology are studied,and a suspicious personnel tracking and recognition system is built.The specific work is as follows:(1)A moving target detection algorithm based on VIBE+ is proposed.The algorithm uses DBSCAN clustering algorithm as a bridge to combine the three-frame difference method with the traditional VIBE algorithm to achieve the complementary advantages.The experiment proves that the VIBE+ algorithm effectively improves the "ghosting" problem of the traditional VIBE algorithm,and avoids the "hole" phenomenon of the three-frame difference method,and meets the real-time requirements of the system in terms of time overhead.(2)A target tracking algorithm based on improved particle filter is proposed.The particle filter algorithm is improved from four aspects: feature enhancement,increasing particle diversity,improving algorithm anti-interference and introducing target template update strategy.The experimental results show that the improved particle filter algorithm effectively improves the robustness of the traditional particle filter algorithm when the environment is moved,the target is occluded or the attitude changes greatly,and the time overhead meets the system requirements.(3)Pedestrian anomaly behavior recognition algorithms were designed.Regional intrusion detection is achieved by performing pedestrian detection in a manually delineated restricted area.The fall behavior of a single pedestrian,by tracking the target pedestrian,obtaining acceleration features and shape features,and constructing rules for fall determination.The energy function based on the motion history map is designed to identify multi-person anomalous behavior.Experiments show that abnormal behavior recognition algorithms have high detection accuracy and time expenditure meets the requirements of real-time system.(4)Finally,based on the above research results,using MFC,Open CV and My SQL technologies,the system architecture design,database design,algorithm modular packaging and user interaction interface design are implemented,and the suspicious personnel tracking and recognition system is realized.