Video Action Recognition System Based On Surveillance Scene

In recent years,the video image automatic identification technology has made significant progress,face recognition,visual technology such as automatic license plate recognition has become People's Daily,has been very widely used,but for most users,video gesture recognition is relatively unfamiliar,this has to do with people's increasing security requirements,So the popularization of video action recognition system based on monitoring scene is more and more urgent.The main work of this paper is as follows:(1)Aiming at the problem of insufficient utilization of bone information by classical spatio-temporal convolutional network(STGCN),this paper designs a residual spatio-temporal graph convolutional network based on multi-information fusion.Firstly,bone edge information and bone point information are used to directly fuse information in high-dimensional space,so that more distinguishable features can be extracted from the model.Secondly,to solve the problem of insufficient nonlinear ability of spatio-temporal convolutional network,residual connection is added to the network,so that the gradient will not disappear when the network deepens.Finally,on the NTU?RGB+D action recognition data set,the accuracy of multi-information fusion residual spatio-temporal graph convolution network is 5.7% higher than top-1of spatio-temporal convolution network.(2)based on simple actions identify needs consideration,this paper designs a residual figure convolution model based on static bone point,through the space diagram convolution and strategies to identify human body static action,on the accuracy of self-built data set reached 98%,can undertake gesture recognition accuracy and stability,in order to achieve the purpose of practical use.(3)Design and implement an action recognition system based on monitoring scenarios,and use the residual graph convolution model based on static skeleton points to effectively and accurately identify actions such as long-term standing(penalty standing),squatting,and long-term raising of hands.It has very practical application value in the fields of inspection by the Disciplinary Committee and prison inspection.