Research And Implementation On Abnormal Human Behavior Detection In Intelligent Monitoring System Of Prison

Along with the development of the society, the improvement of material and spiritualcivilization, people pay more and more attention to safety, so the video surveillance systems havebeen widely used in the subway, bank, hotel, supermarket, community, parking lot, campus andother fields. It plays a more and more important role in the social public security area. Prison is ahighly complex and densely populated place, its management work is not easy, the effective use ofintelligent monitoring can protect the surveillants, prisoners and social security. In the prison,workers care about abnormal behavior by video monitoring, and can receive real-time and reliablewarning sign, this can help the surveillants solving urgent problems quickly and effectively. Sorecognising abnormal behavior accurately and providing timely alarm is the main point of thisthesis.Aiming at foreground extracting and specific target recognition in fixed camera video streams,this thesis studied these problems and implemented the recognition of police and prisoners in prison.Firstly, in order to implement my algorithms later this thesis introduced OpenCV, the Open SourceComputer Vision library, and some basic image processing technology. Secondly, this thesisintroduced some common methods of foreground extracting, especially Gaussian model andCodeBook model. Then compared the advantages with disadvantages of these two methods and animproved algorithm was proposed. Thirdly, this thesis studied several methods of AbnormalBehavior Recognition and the author defines the contour of targets and the Optical flow features ineach labeled region to recognize abnormal event. This thesis used VC++6.0and OpenCV toimplement the algorithm and verified the effectiveness of the algorithm by experiments.