Design And Implementation Of Obstacle Recognition In Train Operation And Equipment Fault Detection System

The railway system is a vital medium for freight transportation in China and dominates the transport sector.In recent years,China’s railway industry has developed rapidly,and the construction of high-speed railways is in full swing.More and more people are choosing trains as a way to travel due to features such as short intervals and no pollution.So much more attention has been given to the safety of train operation.The healthy state of railway trackside equipment and the intrusion of foreign objects on the tracks are important factors that threaten the safety of train operation.Traditional railway obstacle detection relies on manual and non-visual sensor-based methods,which have a limited detection range and are difficult to install and maintain the equipment.Visual sensor-based methods,most of which use traditional image processing methods,are susceptible to environmental factors.Traditional trackside equipment detection using manual and analysis and processing of signal data is costly.It cannot detect defects and faults in equipment appearance in time.Therefore,this thesis proposes an obstacle recognition in train operation and equipment fault detection system.Image processing and deep learning technologies are introduced to realize the recognition of obstacles in train operation and detection of the appearance defects of trackside equipment.Besides,the systematic management of image information is also performed.This thesis researches and analyzes the existing methods of recognizing obstacles in train operation and trackside equipment fault detection,and on this basis designs and implements a recognition of obstacles in train operation and equipment fault detection system.The system contains five functional modules: login and authority authentication,user management,image management,basic information management,and anomaly detection.The whole system is designed based on the MVC model and developed in the form of front-end and back-end separation.The back-end is developed by the Spring Boot framework,and Spring Security is used to authenticate and identify users.The data generated during the system’s work is stored in a My SQL database and cached in Redis caching middleware for specific data information.Anomaly detection is the system’s core module and contains algorithms for obstacle recognition in train operation and equipment fault detection.The obstacle recognition algorithm in train operation is designed and implemented based on YOLOv4 algorithm.To further improve the detection rate,Mobile Net V3 is introduced,and the impact of backbone network change is balanced by SE attention mechanism.The equipment fault detection algorithm introduces RANSAC algorithm and SURF algorithm to match and align images with templates and completes the detection of trackside equipment faults by filtering the matched abnormal areas.The system has passed a number of tests and is running well,which provides platform support for the further research of algorithms for obstacle recognition in train operation and trackside equipment fault detection.