Research And Implementation Of Human Body Tracking In Video Surveillance

In recent years, the technology of net video surveillance and intelligent video surveillance is improving, The occasions which applies video surveillance is more and more.The combination of net and intelligence which is called intelligent net video surveillance promotes the development of security industry, at the same time it greatly reduces the crime rate.The intelligent net video surveillance plays a vital role to protect people and property from the Infringement. A popular research direction in intelligent video surveillance is the human body tracking. This thesis combines the human body tracking in net intelligent video surveillance, detects human body and tracks the human body in client part.The thesis has a certain significance for the the combination of net surveillance and intelligent video surveillance, improves the security of video surveillance monitoring with saving human resources.The main content of this thesis analyses the algorithm of the current human body detection and tracking,improves the algorithm of human body detection and tracking and the tracing efficiency and realizes the human tracking function in client part of surveillance. This thesis firstly describes the background and significance of research, analyses the research status of the human body tracking and the human body tracking in video surveillance system at home and abroad. Then it describes the development of video surveillance system and the function of human body detection and tracking in intelligent video surveillance. Then the human body detection methods at present are classified, also improves detection algorithm of human body, analyzes the human tracking algorithms, an improved algorithm is proposed based on Meanshift algorithm. Finally, realize the algorithm in video surveillance system. The work of this thesis includes the following three points:1.The thesis propose a kind of detection methods that combining hog features and background subtraction method, using the background subtraction method to eliminate the useless video information, improves the detection efficiency, reduces the detection time. At the same time, the human detection method based on hog feature overcomes the defects of the background difference method which can not completely detect the whole human body.2.In order to solve the high real-time requirement and effect of the background in video surveillance, this thesis propose a human body tracing method for video surveillance, firstly using Kalman filter to predict trend of human motion, the using the Meanshift algorithm based on kernel near the predicted location to search the position of human body, last using the Kalman filter to correct the position of the human body. The experimental results show that the improved algorithm proposed in this thesis effectively overcome the background interference caused by irregular shape of human body, have a certain improvement in the human body tracking accuracy and robustness.3.The thesis realizes application of the algorithm in client part of the net video monitoring system. The combination of the net video surveillance and intelligent video surveillance have certain effect to improve the prevention of video surveillance.