Research And Implementation Of Sensitive Behavior Detection Algorithms For Secret Sites Based On Deep Learning

Aiming at the shortcomings of existing video surveillance of secret places,this paper combines the research results of cutting-edge deep learning to study and implement the sensitive behavior detection methods of video surveillance based secret places.This article innovatively uses the deep learning-based target detection technology to detect sensitive targets in secret places,and at the same time to detect sensitive behavior in secret places according to the characteristics of sensitive behaviors,the effect is significant.The research content of this paper is mainly divided into three parts:sensitive target detection in sensitive places,sensitive behavior detection in sensitive places,and intelligent monitoring system design in sensitive places.Firstly,in order to solve the problems of lack of robustness and time consuming of traditional target detection algorithms such as HOG,this paper uses a deep learning-based target detection algorithm YOLOv3,and proposes an improved Secret?YOLO network model.The improved network improves the inference speed and accuracy of the network by merging the BN layer into the convolution layer,using GioU instead of the L2 loss function,and adding a pooling layer,making this network model closer to sensitive target detection in secret places Demand.Secondly,in the part of sensitive behavior detection in secret places,a detection method based on sensitive behavior characteristics is proposed.According to the characteristics of sensitive behaviors,and combined with the results and location of sensitive target detection,this method determines whether it is a sensitive behavior.It can detect four sensitive behaviors such as carrying communication equipment,placing relics,falling,and wandering.Laster.Try adding a pooling layer to the model to increase the depth of the model.The training results show that the accuracy of the model depth can be improved by 2%?4%.Thirdly,in view of the shortcomings of the existing monitoring system in confidential places,this paper designs a set of intelligent monitoring systems in confidential places.The system has functions such as sensitive target detection,sensitive behavior detection,sensitive behavior early warning,sensitive behavior history query,and security evaluation of secret locations.