Research And Implementation Of Manhole Cover Detection And Safety Warning Based On Video Surveillance

As deep learning technology has made rapid progress in the field of computer vision,it has been extensively used in modern society and public infrastructure management.The management of manhole cover has always been a key issue of urban management.With the advent of“smart city" and 5G,cameras has been touching nearly every corner of the city,realizing a faster transmission of data to the server than ever.It has become possible to use video surveillance to manage manhole covers.Under the background,this article researches and implements manhole cover detection and safety warning based on video surveillance.The specific research content is as follows:Improved the YOLOv4 network,which has the best balance between accuracy and detection speed in the single-stage target detection network.This article modified the backbone network structure of YOLOv4,and merged the depth information and the edge information generated based on the Canny operator into the input of the model.Besides,Multi-dimensional Fusion is completed in this paper,which improves the detection effect of ordinary manhole covers and damaged manhole covers.Finally,the effectiveness of the improved manhole cover detection model is proved through comparative experiments.In addition,the improved manhole cover detection model is applied to video detection.And a safety warning mechanism for manhole cover theft and damage detection was proposed.At the same time,a prototype system was designed and implemented.