Design And Implementation Of Escalator Video Security Monitoring System Based On Convolutional Neural Network

With the increasing living standards of people,the number of escalators is increasing.Along with the convenience brought by people,the frequent escalator safety accidents have caused people to pay more attention to the safety of escalators.However,the current safety detection methods for escalators only stay in the traditional detection method.Focusing on the accident when the accident occurs,the corresponding safety switch can be triggered to control the escalator to stop.If an emergency occurs,the pedestrian is required to actively press it.There is no way to ensure that passengers are safe in time to take emergency measures to ensure passenger safety.In order to respond to emergencies in time and protect passengers as much as possible,this paper builds an escalator video security monitoring system based on convolutional neural network.The deep learning algorithm is applied to the escalator video surveillance to judge the posture of the escalator,such as climbing the handrail,squatting,falling,etc.,and then taking corresponding security measures according to the danger level.The work and innovations of this paper are as follows:(1)In view of the poor performance of traditional escalator safety detection methods,this paper proposes a safety detection method based on the dangerous posture of the human body.The improved YOLOv3 algorithm is used to detect the posture of the human body,which improves the detection performance for the safety of the escalator and improves the detection effect in the target occlusion scene.(2)Aiming at the problem of the loss of the intermediate frame target caused by using only the target detection algorithm to process the video,this paper introduces the Deep_SORT multi-target tracking algorithm to establish the association between the frames before and after the video based on the improved YOLOv3 algorithm.It can effectively alleviate and suppress the problem of target "drop frame",and further optimize the recognition effect under the target occlusion scene.(3)Aiming at the actual needs of escalator monitoring,this article designs and implements a video surveillance system with login verification function,attitude detection function and record query function,and reserves the interface to control the running status of the escalator,which can judge the dangerous posture of human body and make response.Finally,the data set is collected and produced in a real scene to test and verify its function.