Research On Climb Detection Based On Deformable Part Model

Video surveillance system plays an increasingly important role in the field of public security.With the increasing numbers of cameras arranged in public places in the city,a vast amount of video data are acquired,how to realize intelligent monitoring has become the focus of current research.And the key point and difficult point in intelligence video surveillance is the analysis and recognition of human behavior.The analysis and recognition of human behavior has been widely used in human-computer interaction and motion analysis technology.The identification of certain behavior such as climbing behavior also has a broad application prospects,where in residential,factory,especially in prisons,warehouses and other special places.Nowadays,the intelligent monitoring system has made some progress,however,due to the postural diversity of human body and the environment variability,human detection based on the video surveillance technology has not achieved the desired results.Two problems have to be solved in Climbing behavior identification,one is the target recognition,check whether there is a human in the picture or video;the other is behavior analysis,the movement characteristics of the human could be extracted,and then identify the behavior of human by comparison model.On the basis of these two points,this paper builds the human climbing identification system,intelligently detect and analysis the climbing behavior of the human,the main achievement are as follows:1.In view of the diversity of human body feature in video,a new method for human detection names multi-decision based on deformable model and color feature is proposed.The main idea is:firstly,use deformable part model for initial testing,then to determine whether the score is greater than the set threshold,if the score is greater than the threshold value and it's not in suspicious range,it can be judged to be the human body;if it's not,the color feature of the target is detected again and the score is regarded as the final result.The experimental results show that the multi-decision detection method based on deformable part model and color feature can improve the detection accuracy in a certain degree.2.The human climbing detection system is developed based on deformable part model and dense trajectory feature,the system is divided into three modules:motion detection module,human detection module and climbing recognition module.The first step of detection is to separate the foreground from the dynamic region in the video,and extract the moving area to be detected,then using the deformable part model for human detection in the motion area.Finally,the human behavior is identified by the use of dense trajectory features.If it's the climb,mark or alarm it,then the intelligent human climbing monitoring is achieved.