Design And Implement Of Substation Intrusion Detection Systerm Based On Machine Vision

Substation is an important hub for connecting power systems and users in social power supply,and its safe and stable operation is of great significance to the entire power system.The purpose of substation intrusion detection is to detect and alarm the intrusion accurately and timely when the substation encounters intrusion problems such as people,cats and dogs,thereby preventing the destruction and theft of the power equipment of the substation,and reducing or even avoiding the loss of the substation due to the invasion,which has a certain practical significance.The thesis applies the research results of the object detection algorithm in the field of machine vision technology to the problem of substation intrusion object detection,and studies and compares the effects of the three object detection algorithms of Faster R-CNN,YOLO and SSD in substation scenarios to select the appropriate algorithm as the substation intrusion detection algorithm of the system.Then,the monocular vision ranging technology is used to complete the safety ranging of the intrusion object and the specific area,and achieve intrusion alarm.The system mainly consists of three parts,namely the server,the image processing terminal and the client.The server mainly completes constructing the database environment,storing the video images and other data and deploying system communication module,wherein the stored data mainly includes the user management information,device management information,alarm images and log information of the substation.The system communication module carried on the server is used for monitoring the connection request and operation instruction of the image processing terminal and the client,parsing the instruction according to the communication protocol and completing the corresponding database operation,etc.,and then replying to the image processing terminal and the client.In addition,the server is also responsible for receiving the alarm push of the image processing terminal,saving the alarm and forwarding it to the client,and the client drives the alarm device to alert and display the alarm information.The image processing terminal acquires the image of the substation site in real time through the secondary development of the camera and performs preprocessing,and then it runs the intrusion object detection algorithm of the system to analyze and process the preprocessed image frame by frame to determine whether there is an intrusion,if any,then the alarm will be pushed to the server,and the received alarm information is saved and forwarded by the server.The client is used for front-end human interaction,mainly responsible for user login,management of user information and management of device information,real-time preview of monitoring videoof substation and historical playback,moreover,it also has the function of driving the alarm device to alert and display alarm information after receiving the alarm information forwarded by the server.The substation intrusion detection system based on machine vision technology designed in the thesis realizes the accurate detection and hierarchical alarm of invading object including people,cats and dogs in substation environment.Experiments show that the average accuracy of substation intrusion detection system of three intrusion objects of people,cats and dogs in the substation scenario is 90.8%,and the detection speed can reach 32 fps on the GPU1080 Ti machine,which has the characteristics of high accuracy,good real-time performance,and low cost,and high stability and so on.Therefore,the whole system has certain significance for ensuring the safe and stable operation of the substation.