摘要
为解决铁路站场接触网点云噪声分布不规律及语义分割难度大的问题,提出一种智能提取方法,以增强接触网异常检测能力.首先,对站场接触网场景数据进行深入分析,构建导线及钢轨顶面点云提取的知识框架;其次,考虑站场接触网点云空间特征,设计站场关键要素点云的分割与融合滤波方法;然后,建立站场接触网强空间语义约束规则,提出知识引导的导线特征智能精细提取方法;基于此,采用WHU-TLS等站场点云数据集,搭建实验平台并开展实验分析,实验结果表明:在部分点云缺失以及噪声干扰等复杂环境下,本文方法易于操作且自动化程度高,相比传统导线特征提取方法耗时最少,100 m范围内站场接触网导线特征提取的平均精度达到±5 mm,能够有效支撑铁路站场接触网几何特征的智能检测.
To address the irregular noise distribution and difficult semantic segmentation in overhead catenary point clouds of railway stations and enhance the detection of overhead catenary anomalies,an intelligent extraction method was proposed.Firstly,the overhead catenary scene data of railway stations was analyzed,and the knowledge framework for wire and rail top surface point cloud extraction was constructed.Secondly,the spatial features of overhead catenary point clouds of railway stations were considered,and the segmentation and fusion filtering methods of key element point clouds of railway stations were designed.Then,the strong spatial semantic constraint rules of the overhead catenary of railway stations were established,and a knowledge-guided intelligent and fine extraction method for wire features was proposed.On this basis,WHU-TLS and other station point cloud datasets were used to build an experimental platform and carry out experimental analysis.The results show that in complex environments with partial missing of point clouds and noise interference,the proposed method is easy to operate and highly automated.Compared to traditional methods for extracting wire features,it requires the least time and achieves an average precision of±5 mm in extracting overhead catenary wire features within 100 m,effectively supporting the intelligent detection of geometric features in overhead catenary of railway stations.
作者
朱军
张传军
赵剑峰
王学柱
付林
黄智勇
郭鹏飞
ZHU Jun;ZHANG Chuanjun;ZHAO Jianfeng;WANG Xuezhu;FU Lin;HUANG Zhiyong;GUO Pengfei(Faculty of Geosiences and Engineering,Southwest Jiaotong University,Chengdu 611756,China;Department of Power Supply,China Railway Beijing Group Co.,Ltd.,Beijing 100860,China;School of Civil Engineering and Geomatics,Southwest Petroleum University,Chengdu 610500,China;Xikang Headquarter of Xi’an–Chengdu Passenger Dedicated Line Co.,Ltd.,Xi’an 710000,China;Institute of Computing Technology,China Academy of Railway Sciences Corporation Limited,Beijing 100081,China)
出处
《西南交通大学学报》
北大核心
2025年第6期1333-1341,共9页
Journal of Southwest Jiaotong University
基金
国家自然科学基金项目(42271424)。
关键词
激光点云
铁路站场接触网
知识引导
导线特征
智能提取
LiDAR point cloud
overhead catenary of railway station
knowledge guide
wire feature
intelligent extraction
