摘要
为降低传统手工检测的工作强度,提高检测效率与精度,将结构光三维成像技术应用于钢轨检测。由于铁路现场扫描的钢轨点云数据量庞大,且存在大量噪声点与离群点,若不进行精简,对后续点云识别与配准的效率与精度会产生较大影响。为解决上述问题,首先运用最大主成分方向、法线方向和质心对原始点云进行摆正;然后根据空间位置与密度聚类结果除去离群点,根据钢轨廓形信息执行直通滤波,除去轨腰点云中粘连的噪声点;最后将钢轨点云分割为水平平面点云、竖直平面点云以及曲面点云,针对不同位置的点云,基于点距设置不同的体素采样参数进行精简。铁路现场试验表明,该方法可以有效去除钢轨点云中的噪声点与离群点,精简比为80%左右,最大程度保留原始点云的形貌特征与细节信息的同时,点云配准效率提高近4倍,为高效精确的点云处理奠定基础。
In order to reduce the work intensity of traditional manual detection and improve detection efficiency and accuracy, three-dimensional imaging technology of structure light is applied to rail detection. The data of rail point cloud collected from the railway site is huge with large number of noise points and outliers, which will have a significant impact on the efficiency and accuracy of subsequent point cloud identification and registration if the data is not simplified. This paper proposes a novel method to solve above problems. Firstly, the direction of the maximum principal component, the normal direction, as well as the centroid are used to straighten the original point cloud. The outliers are then removed according to the spatial location and density clustering results. After that, a pass filter is performed according to the rail profile information to remove the adherent noise points in the rail web point cloud. Finally, the rail point cloud is divided into horizontal plane point cloud, vertical plane point cloud and curved surface point cloud. For point clouds at different locations, different voxel sampling parameters are set based on the point distance to conduct the simplification. The railway field experiments show that this method can effectively remove 80% of noise points and outliers from the rail point cloud. The registration efficiency of the point cloud is improved by almost 4 times, while the morphological characteristics and detail information of the original point cloud are preserved to the greatest extent, which lays the foundation for more efficient and accurate point cloud processing.
作者
税文
王培俊
屈仁飞
赵瑞
赖宸宇
罗鑫
SHUI Wen;WANG Peijun;Qu Renfei;ZHAO Rui;LAI Chenyu;LUO Xin(School of Mechanical Engineering,Southwest Jiaotong University,Chengdu 610031,China)
出处
《铁道标准设计》
北大核心
2022年第7期36-41,共6页
Railway Standard Design
基金
四川省科技计划重点研发项目(2019YFG0046)。
关键词
钢轨检测
点云精简
点云摆正
直通滤波
平面分割
分层滤波
rail detection
point cloud simplification
point cloud straightening
pass filter
plane segmentation
stratified filter
