铁路场景三维点云分割与分类识别算法

发布时间：2018-02-04 06:13

  本文关键词： 三维点云 区域生长 分割 视点特征直方图 分类识别　出处：《仪器仪表学报》2017年09期 　论文类型：期刊论文

【摘要】：铁路限界侵入检测对保障高速铁路安全具有重要意义,基于激光三维点云分割与分类识别的异物侵入检测具有准确、直观的优点,在诸如隧道口和站台的铁路重点区域监测中具有广泛应用前景。设计了一种带动二维激光雷达进行俯仰运动的装置用于铁路三维点云的采集,基于法线方向一致性原则提出采用区域生长分割算法解决欧氏聚类分割和随机采样一致性(RANSAC)分割造成的过分割和欠分割问题;针对分割后的单物体点云,提出利用视点特征直方图(VFH)进行不同目标的三维点云特征提取,基于不同物体VFH建立KD树,并利用最近点搜索方法完成单物体点云分类识别。铁路场景典型物体的分类实验结果表明,本算法对铁路场景典型物体的分类识别准确率大于90%。
[Abstract]:Railway boundary intrusion detection is of great significance to ensure the safety of high-speed railway. The foreign body intrusion detection based on laser 3D point cloud segmentation and classification recognition has the advantages of accuracy and intuition. It has a wide application prospect in the monitoring of railway key area such as tunnel entrance and platform. A kind of device which can drive 2-D lidar to carry on pitching motion is designed for railway 3D point cloud acquisition. Based on the principle of normal-direction consistency, a region growing segmentation algorithm is proposed to solve the problems of over-segmentation and under-segmentation caused by Euclidean clustering segmentation and random sampling consistency segmentation. Aiming at the point cloud of a single object after segmentation, this paper presents a method of extracting 3D point cloud features from different objects by using view feature histogram (VFH), and builds KD tree based on different objects VFH. The experimental results show that the accuracy of this algorithm is more than 90% for typical objects in railway scene.
【作者单位】： 北京交通大学机械与电子控制工程学院;北京控制工程研究所;
【基金】：国家重点研发计划(2016YFB1200100)项目资助
【分类号】：TP391.41;U298
【正文快照】： 0引言侵入铁路限界的物体严重影响运营安全,在车站站台、咽喉区、隧道口、通道门等重点区域,识别侵入物体的具体位置、大小和类别具有重要意义。以防护网为代表的接触式检测方法无法获知物体的大小进而评估其危害,非接触式检测中的视频分析方法会受到天气及光线条件的影响[1],，

本文编号：1489559