Detection and analysis of key component of train's locomotive based on image analysis techniques
Detection and analysis of key component of train's locomotive based on image analysis techniques
复制标题
基于图像分析技术的列车机车关键部件检测与分析
DOI：
10.1109/iwc.2016.8068383
复制
发表时间：
2016
期刊：
影响因子：
--
通讯作者：
Zhiming Liu
中科院分区：
文献类型：
--
作者：
Yunjie Zhong;Xiaorong Gao;Jianping Peng;Lin Luo;Zhiming Liu
随着我国高速铁路的快速发展，机车的运行安全问题引起了铁路部门的高度重视。机车车底作为机车安全的一个重要检测环节，存在诸多不安全因素，其中走行部故障是影响机车安全运行的最直接因素。传统的检测方法效率极低，人工无法进行轨旁动态检测。为了提高检测的有效性，同时能够对机车底部关键部件进行动态检测，研究有效的机车底部故障检测与识别方案具有重要意义。在机车底部故障检测与识别系统中，主要研究了基于图像比较法的走行部变形、损伤、遗漏、有无异物等故障的检测。提高检测率和降低误检率是本课题的主要目的。采用多特征融合方法对走行部检测数据进行处理，将图像边缘检测和视觉相似性方法相结合，综合考虑图像的边缘、对比度、亮度、结构等特征，提高了检测率，降低了误检率。
With the fast development of Chinese high-speed railway, the operation safety of locomotives has attracted great attention of railway departments. As an important detection aspect of locomotive safety, the locomotive bottom has lot of insecurity factors, and faults of running gear are the most direct factor affecting the safe operation of locomotives. Generally, the traditional method is extremely inefficient, and manpower cannot allow wayside dynamic detection. In order to improve the effectiveness of detection, and at the same time allow dynamic detection of key components at the locomotive bottom, the research on effective detection and identification schemes of locomotive bottom faults is promising. As for the whole detection and recognition system of locomotive bottom faults, this paper mainly studies detection of running gear deformation, damage, omission, and the presence of foreign objects and other faults based on the image comparison method. Increasing the detection rate and reducing the false detection rate are the main purpose of this project. This paper used multi-feature fusion method for process the detection data of the running gear, which combined the image edge detection and visual similarity method, comprehensively considered the edges, contrast, brightness, and structural characteristics of the image, finally improved the detection rate and reduced the false detection rate.