机器视觉在轨道扣件检测上的应用综述  

Overview of the application of machine vision on the test of rail fasteners

作　　者：黄石甫 曹广如[1] 李斌[1] 程志[1] HUANG Shipu;CAO Guangru;LI Bin;CHENG Zhi(Zhuzhou Times New Material Technology Co.,Ltd.,Zhuzhou,Hunan 412007,China)

机构地区：[1]株洲时代新材料科技股份有限公司,湖南株洲412007

出　　处：《轨道交通材料》2024年第3期45-53,共9页MATERIALS FOR RAIL TRANSPORTATION SYSTEM

摘　　要：机器视觉通过模拟人眼感知,对图像进行采集、预处理、特征提取、识别定位等步骤,结合深度学习与图像处理算法,实现轨道扣件的高精度、实时、自动化检测。具体应用包括:运用边缘检测、模板匹配识别扣件缺失、松动;使用机器学习分类识别磨损、裂纹等表面缺陷;采用三维视觉测量扣件几何参数偏差,显著提升了检测精度、范围和效率。针对机器视觉在轨道扣件检测方面的一些研究应用,从传统的视觉检测方法出发,扩展到基于深度学习的检测方法,阐述视觉检测基本原理、关键技术和实际应用效果,以及面临的挑战与未来发展趋势。By means of simulating human eye perception,machine vision performs image acquisition,pre-processing,feature extraction,identification and positioning,and combines deep learning and image processing algorithms to realize high-precision,real-time and automatic detection of track fasteners.Specific applications include:using edge detection,template matching to identify missing or loose fasteners;using machine learning to classify and identify surface defects such as wear and cracks;The application of 3D vision to measure geometric parameter deviations of fasteners significantly improves detection accuracy,range,and efficiency.Aiming at the research and application of machine vision in the detection of track fasteners,this paper starts from the traditional visual inspection methods,then expands to the inspection methods based on deep learning,and expounds the basic principle,key technologies and practical application effects of visual inspection,as well as the challenges and future development trend.

关 键 词：机器视觉 轨道扣件 图像数据 深度学习 

分 类 号：U270.7[机械工程—车辆工程]