Machine Vision Intelligent Detection Technology For Foreign Body Intrusion In Railway Perimeter

As an important pillar of passenger and freight transportation and the lifeline of the national economy,railways carry more than 60% of passenger transportation activities.It is particularly important to ensure the safe operation of railways.Especially for ordinary-speed railways,excluding some improper railway operation management and operation,one of the other reasons that may cause serious traffic accidents is the intrusion of foreign objects in the perimeter of the railway.Despite the continuous improvement of my country’s railway infrastructure,the intrusion of various foreign objects still occurs frequently.Foreign objects are random,sudden and unpredictable.Relying on traditional manual inspection methods along the line has large workload,low inspection efficiency,and the risk of missed inspections,which cannot meet the needs of railway safety operation and management;traditional contact detection(such as building a catenary)is relatively mature and has a low false detection rate,but cannot obtain detailed information on foreign bodies.In order to realize the all-weather real-time monitoring of railway foreign body intrusion and meet the requirements of intelligent railway safety operation,it is urgent to design and develop a real-time,reliable and accurate automatic foreign body intrusion alarm monitoring system.In view of the shortcomings of traditional methods,such as high cost,low detection accuracy,and the risk of missed detection,this paper proposes a foreign object detection scheme based on machine vision.To achieve foreign object tracking and trajectory prediction,a set of railway foreign object intrusion intelligent detection system is developed.The effectiveness of the system is verified by the application demonstration in the actual railway scene and the evaluation of the accuracy of the results.The main research contents and corresponding results of this paper are as follows:(1)In order to meet the real-time requirements of railway foreign body intrusion detection,improve the detection accuracy and efficiency,and greatly reduce the probability of missed detection and false detection,a target detection scheme based on deep learning is determined.At the same time,by comparing the target detection accuracy of the one-stage YOLOv5 and the two-stage Faster R-CNN convolutional neural network model in the actual complex railway scene,and comprehensively comparing the experimental results,a highefficiency scheme suitable for foreign object detection in the railway scene is determined.(2)Aiming at the problems of low target tracking and positioning accuracy and identity change,a simple online-based deep correlation metric tracking algorithm Deep SORT is proposed,which greatly solves the problems of traditional single target detection and tracking target occlusion loss and weak data correlation.By establishing the deep correlation between the frames before and after the target,the problem of the multi-target tracking algorithm in the switching between new and old targets and ID identification is solved.At the same time,in order to solve the problem of insufficient predictability of potential risks in the process of system detection,a Kalman filter method is proposed to realize the trajectory prediction of pedestrian targets,and combined with the ROI area to determine the trend of intrusion,so as to improve the accuracy of system detection and early warning.(3)Integrate target detection and recognition algorithms and multi-target tracking algorithms,and design and develop a prototype system for intelligent detection of foreign body intrusion in railway perimeter based on platform-side and mobile-side monitoring equipment.The prototype system consists of two parts: the platform end and the mobile end: the platform end is a set of UI software interface designed by Py Qt based on the server base station,and embedded various functional modules;the mobile end is mainly responsible for data collection and data transmission.Finally,the actual test of the system is carried out,and the detection and early warning effect is good.This has great research value for realizing an automated,intelligent and highly reliable railway foreign body monitoring platform.