Research On Pedestrian Detection And Tracking In Intelligent Video Surveillance

Intelligent video surveillance is an important research direction in the field of computer vision, its wide range of applications, such as many aspects of security, traffic management, banks, hospitals, etc., has important significance. It is through the analysis of video sequences for detection, identification and tracking of objects in the scene to be monitored and found to provide a basis so that people can make the early warning and rapid response to abnormal behavior behind the target for behavior recognition and events. Intelligent video surveillance system in which people are most concerned about pedestrian detection and tracking algorithms research has important significance. This paper analyzes the common understanding of the target detection and tracking method for pedestrian target under different lighting, occlusion detection and tracking conditions were studied. The main contents are as follows:1) After a detailed study of the human skin color detection algorithm, propose a face detection method based on YCbCr cluster map, enhanced color detection of complex environmental robustness.2) Introduce the compressed sensing algorithm, and improve the tracking algorithm based on compressed sensing in two ways, on the one hand is the target feature extraction, get another complementary characteristics on the basis of the original features, and then combined using the method described multi-feature object and background, and enhance the stability of the target; the other is to add detection feedback, use the results of detection to correct the drift target. Parallel processing is introduced to improve tracking algorithm based on compressed sensing technology with OpenMP; and through experiments prove the effectiveness of the algorithm under different circumstances.3) Study the problem of video tracking targets occlusion treatment method is proposed based on the color histogram of particle filter algorithms, and experimentally verify the robustness of the algorithm block processing. 4) Different members of the laboratory research group studied the different techniques in intelligent video surveillance area, together constitute a set of intelligent video surveillance systems, including the prospects for detection and segmentation, pedestrian detection and tracking, human behavior recognition, hovering stay detection, abandoned object detection, abnormal events found and so on, this paper is completed pedestrian detection and tracking section, which is the basis of other modules, the proposed algorithm is good or bad will indirectly affect the performance of the whole system.