轨道故障动态三维检测与识别系统

Dynamic 3D Track Fault Detection and Identification System

轨道线路作为城市轨道交通不可或缺的基础设施,其关键部件状态的动态检测、故障的自动识别与列车行车安全息息相关。针对现有2D检测手段存在的检测精度低、检测功能单一、易受环境影响、人工效率低等问题,研制了一套轨道故障动态三维检测与识别系统。该系统基于三维成像技术获取钢轨、枕木和扣件的二维图像及深度信息,结合阈值分隔技术和YOLOv4目标检测算法精确定位检测区域,利用基于深度学习的三维点云匹配算法识别检测区域状态,并在数据终端实时显示检测与识别结果。经试验验证,该系统对钢轨磨耗的检测精度达到±0.25mm,轨表、枕木、扣件的故障识别率达96.7%以上,且能排除环境因素的干扰,可用于实时检测轨道线路多种故障,有助于轨道检修作业的智能化升级。

As the rail tracks constitute the indispensable infrastructure of urban rail transit,the dynamic detection of states and the automatic identification of faults in their key components are closely linked to the operation and safety of the train.To solve the problems of the existing 2D detection methods,such as low precision of detection,single function detection,susceptibility to the environment,and low labor efficiency,a dynamic 3D track fault detection and identification system is developed.Based on the 3D imaging technology,this system can obtain the 2D images and depth information of steel rails,sleepers and fasteners,combine the threshold separation technology and the YOLOv4 object detection algorithm to accurately locate the detection area,use a deep learning-based 3D point cloud matching algorithm to identify the states of the detection area,and display the detection and identification results in real time on the data terminal.It’s verified by tests that the rail wear detection precision of this system can reach±0.25 mm and its fault identification ratio for rail surfaces,sleepers and fasteners is above 96.7%.It can also suppress the interference of environmental factors.This system can be used for the real-time detection of multiple faults in rail tracks and is helpful for the intelligent upgrading of track maintenance work.