首页|基于点云特征的城市道路标识线提取与分类

基于点云特征的城市道路标识线提取与分类

扫码查看
原文链接
  • 万方数据
  • 维普
为了解决基于车载激光雷达(LiDAR)点云数据中的道路标识线提取完整度与提取精确度方面数值偏低等问题,提出了一种基于点云多元特征的道路标识线快速提取方法.在城市道路标识线的强度信息、几何信息和语义信息基础上,结合路面点云的强度特征、高程特征和点密度特征,生成多个地理参考图像,对多元特征图像进行特征提取与填充,再利用Ostu算法以及Alpha shapes算法实现道路标识线点云精提取,并根据标识线的几何、语义信息和模型匹配方案实现标识线的细分类,进行了理论分析和实验验证,取得了澳大利亚某城市道路的点云数据.结果表明,提取的短虚线、斑马线、单向转向箭头、长虚线的准确率均高于96％,召回率均达到91％及以上,综合评价指标均达到94％及以上.这些结果对无人驾驶领域研究起到了添砖加瓦的作用,也为城市数字化建设提供了一定的参考价值.
Extraction and classification of urban road marking lines based on point cloud features
In order to solve the problem of the low value of road marking line extraction integrity and accuracy based on vehicle-mounted light detection and ranging(LiDAR)point cloud data,a fast road marking line extraction method based on multiple features of the point cloud was proposed.Based on the strength information,geometric information,and semantic information of urban road marking lines,combined with the strength feature,elevation feature,and point density feature of the road surface point cloud,multiple geographic reference images were generated,and the feature extraction and filling of the multiple feature images were carried out,and then the Ostu algorithm and Alpha shapes algorithm were used to achieve the precise extraction of the road marking line point cloud.According to the geometric and semantic information of the marking line and the model matching scheme,the fine classification of the marking line was realized.The theoretical analysis and experimental verification were carried out,and the point cloud data of a city road in Australia was obtained.The results show that the accuracy of the extracted short dotted line,zebra line,one-way steering arrow and,the long dotted line is higher than 96％,the recall rate is 91％and above,and the comprehensive evaluation index is 94％and above.The study has contributed to the research in the field of driverless driving and also provided certain reference values for the construction of urban digital.

laser techniquemarking line extractionmultiple characteristicsvehicle-mounted light detection and ranging point cloudbinarizationiterative closest point template matching

郑帅锋、王山东、张陈意、王伦炜

展开 >

河海大学地球科学与工程学院,南京 211100,中国

激光技术 标识线提取 多元特征 车载激光雷达点云 二值化 临近点迭代模板匹配

江苏省幸福河湖评价标准和骨干河道管护评估标准应用研究资助项目

JSZC-320000-HYGS-C2021-0156

2024

10.7510/jgjs.issn.1001-3806.2024.01.005

激光技术
西南技术物理研究所

激光技术

CSTPCD北大核心
影响因子：0.786
ISSN：1001-3806
年,卷(期)：2024.(1)
  • 9