Design And Implementation Of Transmission Tower Anomaly Recognition And Early Warning System Based On Deep Learning

With the rapid development of urban construction,the destruction of transmission lines has been increasing in recent years.Large-scale power outages caused by insulators falling off,power grid foreign bodies and external force damage have occurred.To repair these accidents,Power grid companies spend a lot of manpower and resources on these every year.Since AlexNet won the ImageNet classification competition in 2012,deep learning has developed like an explosion in the image field.In recent years,more and more recognition tasks in the image field have achieved better performance in deep learning solutions.This paper designs and develops a transmission tower anomaly recognition and early warning system based on deep learning,which can timely and accurately identify insulator missing,grid foreign bodies and external force damage in the power grid.The system includes cruise master stations,external force damage preventing devices and a master station.The cruise master stations are server hosts,in whice the insulator missing and the power grid foreign bodies are identified from the aerial video;the external force damage preventing devices are embedded device equipped with a camera working on the transmission iron frame,and the construction vehicles near the transmission line is identified from the camera.These two systems aggregate the anomaly and alarm information to the master station and promptly notify the user to take appropriate action.The image recognition of this system uses YOLOv2 object detection algorithm,and the key frame extraction algorithm based on dynamic background update is used to extract key frames from the original video for detection.This paper gives a detailed description of the design and development of the transmission tower anomaly recognition and early warning system from the system proposal,system requirements analysis,system architecture and detailed design,algorithm implementation,hardware and software system implementation and system operation test.