Research And Implement Of Human Abnormal Activity Recognition In Intelligent Video Surveillance

In recent years, with the increasing demand for security automation System andreal-time monitoring system in the area such as banking, electric power, transportation,security and military facilities,video surveillance is growing rapidly and has beenapplied in wide areas, study on this field also causes more and more concerns ofresearch institutions and scholars. This paper presents a thorough study of theimplement of human activity recognition in intelligent video surveillance based on theresearch of moving targets detection, shadow suppression and moving target tracking.On the research of moving targets detection,we propose an improved algorithmbased on gaussian mixture model for targets detection,this improved algorithm reducesinterference of outside factors and improves the accuracy of targets detection. At last weadopt a method based on shadow properties in HSV color space to suppress the shadow.On the research of moving objects tracking, we take the center of mass,width,height and movement speed of the target as essential features, after the locationinformation of moving target is acquired,we take it as the initialization parameters ofKalman Filter.At last we forecast the location of moving target,then match and tarck theresult.On the research of human abnormal activity recognition, state space method isapplied in this paper, human continuous motion is decomposed into an image sequencebased on time.Then we recognize the moving human bodies activity based on HiddenMarkov Model.Through the implement of intelligent video surveillance and a lot of expriments,wehave showed that the algorithm in this paper is very reliable and the function of humanabnormal activity recognition in the system has very high accuracy rate..