Abnormal Behavior Detection Based On Indoor Scene And System Implementation

Since the beginning of the twenty-first century, high–tech are used more and morewidely, and as the central part of the present social security system, monitoring thecrowd and in-kind are more and more intelligent. At present, intelligent monitoring hasalready been an important and focus research filed of the industry because of theworldwide technical staff of the monitoring field study and innovate the monitoringtechnology constantly. Forward development of the related monitoring technology alsocontributed to the construction of the world within a matter of national security, such asfinance, transportation, defense and other important areas of real-time, accuracy andintelligent monitoring.Based on customer demand required in the interior scenes of abnormal behaviorunder control, the paper studied some more common target tracking algorithms andabnormal behavior detection algorithms under this scenes, and then the whole systemmainly divided into the target tracking module and abnormal behavior detection(identification) module. In the target tracking module, based on less computation anddelay, slow template updating and high misjudgement of Mean Shift algorithm andquick template updating and high accuracy rate, much computation, bad particlediversity of Particle Filter, the author puts forward the target tracking algorithm basedon Kalman filtering method (Kalman) to solve the above algorithms deficiency. In theabnormal behavior detection (identification) module, the authors put forward theHidden Markov Model algorithm to achieve the function of this module.The experimentproved the effect of this method of abnormal behavior detection is good.This thesis analyses the disadvantage of intelligent monitoring system facing thecommon and special scenes, and then put forward the technology which is appropriatewith itself. Specifically, in view of the target tracking module and abnormal behavioridentification module the author put forward target tracking algorithm based on KalmanFiltering method and behavior identifying algorithm based on Hidden Markov Modelmeanwhile realizing the two modules of the system. The project team including theauthor according to their proposed tracking and abnormal behavior recognizing algorithm, combined with the open source library OpenCV2.0develop an intelligentmonitoring platform based on indoor and outdoor scenes. The platform has goodproperty in some respect with real–time, false detecting rate.