摘要
研究基于机器视觉的铁路货车钩舌状态自动化检测方法,为实现钩舌销的高精度识别,保证铁路货车安全运行,设计了基于机器视觉的铁路货车钩舌状态自动化检测方法。从多个视角采集铁路货车钩舌图像,利用模板匹配算法将定位钩舌目标,获得感兴趣区域图像,提取LBP、HOG、Harr-like特征,采用核主成分分析法对多特征作降维处理,采用BP-AdaBoost建立钩舌状态检测模型,实现钩舌状态检测。实验结果表明,所研究方法可实现铁路货车销孔的精准检测,检测效果的F1score、AUC、G-mean均优于其他方法的相应数值。
In order to achieve high-precision identification of hook tongue pins and ensure the safe operation of railway freight cars,a machine vision based automatic detection method for the hook tongue status of railway freight cars was designed.Collect railway freight car hook tongue images from multiple perspectives,use template matching algorithm to locate hook tongue targets,obtain region of interest images,extract LBP,HOG and Harr-like features,use kernel principal component analysis to reduce the dimensionality of multiple features,and use BP-AdaBoost to establish a hook tongue state detection model to achieve hook tongue state detection.The experimental results show that the proposed method can achieve precise detection of railway freight car sales holes,and the detection results of F1Score,AUC,G-mean are superior to those of other methods.
作者
徐建喜
吴迪
谭勇
石怀银
XU Jianxi;WU Di;TAN Yong;SHI Huaiyin(China Energy Railway Equipment Co.,Ltd.,Beijing 100120,China;Mechanical Power Design and Research Institute,China Railway Engineering Design Consulting Group Co.,Ltd.,Beijing 100071,China;Shenyang Huixin Mechanical&Electrical Equipment Manufacturing Co.,Ltd.,Shengyang 110000,China)
出处
《电子设计工程》
2025年第5期86-89,95,共5页
Electronic Design Engineering
