基于三维点云的铁路接触网支柱定位检测研究

周新宇

摘 要

随着经济的快速发展，我国的铁路建设进程也迈入了新的发展阶段。接触网是铁路的重要组成部分，做好接触网的施工质量控制是保证铁路稳定运行的重要手段，而提高检测效率是有效方法之一。支柱作为接触网的重要组成部分，其定位的准确性极大的影响了其它接触网构件的安装工作，而目前常用的定位检测方法主要是人工逐个测量判断，面对复杂的铁路建设环境，此方法存在检测效率低的问题，因此有必要为支柱定位检测探寻一种相对高效的新方法。

在对国内外相关研究和实际应用现状梳理的基础上，提出了一种基于三维点云的支柱定位检测新方法。首先根据相关的标准规范，确定了定位检测指标为支柱限界，并明确了定位检测的安全范围确定方式；然后提出了利用无人机技术来获取铁路图像，通过软件处理图像来获取铁路场景的三维点云数据；为了能得到铁轨和支柱的单体点云集，提出结合二次去噪进行预处理、利用基于法线的点云分割以及基于扩展快速点特征直方图的点云描述和识别；并提出了距离计算和结果判断算法，可以对识别出的支柱和铁轨进行最终的定位检测处理，并将结果可视化呈现。最后通过对一段模拟铁路的实证分析和误差分析，验证新方法的可行性。

研究提出的非接触式可视化支柱定位检测方法，可以帮助现场的检测人员快速直观的确认支柱安装位置的准确性，及时提醒工作人员核查潜在的不达标支柱，丰富了铁路施工现场支柱定位检测方式，也为三维点云和点云处理算法在铁路检测方面的应用提供了新思路。

关键词：铁路接触网；支柱定位检测；三维点云；可视化；距离算法

Abstract

With the rapid economic development, China's railway construction process has entered a new stage of development. The overhead contact system is an important part of the railway, and the construction quality control of the overhead contact system project is an important means to ensure the stable operation of the railway, and improving the detection efficiency is one of the effective methods. The pillar is one of the important components of the overhead contact system. Its positioning accuracy greatly affects the installation work of other catenary components. At present, the commonly used positioning detection methods are mainly manual measurement and judgment. There is a problem of low detection efficiency, so it is necessary to explore a relatively efficient new method for pillar positioning detection.

This paper summarized the current domestic and international researches and actual application status, and proposed a new method for pillar positioning detection based on 3D point cloud. First, according to the relevant standard specifications, the positioning detection index was determined as the pillar limit, and the method for determining the safety range of the positioning detection was clarified; then a method using Unmanned Aerial Vehicle technology to obtain railway images and processing the images through software to obtain three-dimensional railway scene Point cloud data was proposed; In order to obtain the single point cloud collection of rails and pillars, a combination of secondary denoising for preprocessing, point cloud segmentation based on normal and point cloud description and recognition based on extended fast point feature histogram were proposed; and distance calculation and result judgment algorithms were proposed, it can perform the final positioning detection process on the identified pillars and rails, and visualize the results. Finally, through the empirical analysis and error analysis of a section of simulated railway, the feasibility of the new method was verified.

The non-contact visual pillar positioning detection method proposed by the study can help the on-site inspectors to quickly and intuitively confirm the accuracy of the pillar installation position, promptly remind the staff to check the potential substandard pillars, and enrich the railway construction site pillar positioning detection methods. It also provides new ideas for the application of 3D point cloud and point cloud processing algorithms in railway inspection.

Keywords：Overhead contact system of railway, Pillar positioning detection, 3D point cloud, Visualization, Distance algorithm