Research And Implementation Of Suspicious Behavior Detection Technology In Warehouse Area

Suspicious behavior detection technology is a key technology for video surveillance in large venues.This technology turns the video surveillance system from the previous look-and-see analysis to an early warning or even an early warning.This technology has greatly reduced the workload of video surveillance personnel while significantly improving the response speed of large-scale venues to respond to emergencies.In recent years,although the crowd suspicious behavior detection technology has made great progress,there are few studies on the suspicious behavior of the crowd in this particular scenario of the warehouse area.Therefore,it is of great theoretical and practical significance to conduct in-depth research on the suspicious behavior detection technology of the crowd in this specific scenario of the warehouse area.To this end,this thesis studies the key technologies such as the crowd representation method and the suspicious behavior discrimination method for the suspicious behavior detection technology in the warehouse area.The main work and innovations of this round include the following aspects:First,for the specific scenario of the warehouse area,the block energy map is used to analyze the suspicious behavior of the crowd.At the same time,it also introduces the representation method and rationality of the block energy map.Second,the target tracking algorithm based on SIFT and YOLO is used to reasonably divide the blocks of the entire scene,which solves the problem that the evenly divided blocks are affected by the perspective effect.Third,after experimental screening of multiple neural networks,a radial basis function(RBF)neural network was finally selected to process the data transformed by the block energy map.Finally,after experimental verification,the target tracking algorithm is used to reasonably divide the scene generation block energy map.And the generated data is brought into the radial basis function(RBF)neural network and the good results of the suspicious behavior of the five populations defined in this paper have been achieved.