Moving Human Detection And Abnormal Behavior Analysis In Video Surveillance

Intelligent video surveillance introduces the technology of computer vision and patternrecognition technology in the video surveillance, which is taking more and more attention. Theanalysis of human abnormal behavior in video is an increasingly important research direction ofintelligent video surveillance system. It can achieve intelligent detection of human abnormalbehavior by using the technology of computer vision, complete security tasks efficiently, and save alot of human and material resources. The article is mainly for moving human abnormal behavioranalysis in video, which is specifically related to moving human body detection, tracking andabnormal behavior detection.In the moving objects detection, in order to extract the foreground target, the paper proposesbinary mask background model method..By continuously updating the background template whenthe total of0is in a small value, the initial background is established. Then, using backgroundsubtraction method extracts moving objects. At the same time, the background is updated adaptivelyduring objects detection in order to reduce the impact of external factors.In objects tracking, the paper gives a tracking method of adjacent frames area match andgrayscale match. In the binary adjacent frames, moving target owns a lot of overlap area to achieveobjects matching and tracking. If objects merge or split, using Kaman filtering and grayscalehistogram matching method tracks objects.In the analysis of abnormal behavior, the paper analyzes by extracting human features (humancenter and length-width ration) and human-moving-trajectory to the detection of abnormal behavior.The rectangle length-width ratio is taken as the criteria of judging the body fall. The hoveringbehavior is detected based on the direction angle change of movement trajectory. In the remnantsdetection, the system analyses the relation of blobs, and gets the moving state of related blobs,accordingly judges that the blob belongs to remnants.The paper uses Visual C++6.0development tools and OpenCV computer vision library toverify the performance of the algorithm. The experimental results show that the detection algorithmcan quickly, accurately achieve the identification of human movement abnormal behavior in thevideo.