测绘通报 ›› 2023, Vol. 0 ›› Issue (5): 67-71,139.doi: 10.13474/j.cnki.11-2246.2023.0138

• 学术研究 • 上一篇    下一篇

融合车载LiDAR点云和全景影像的道路标线自动提取方法

牛鹏涛1,2, 曹毅3,4, 张恩朝5, 漆洋1   

  1. 1. 成都理工大学地球科学学院, 四川 成都 610059;
    2. 河南工业职业技术学院, 河南 南阳 473009;
    3. 中国石油化工股份有限公司西北油田分公司, 新疆 乌鲁木齐 830011;
    4. 河海大学地球科学与工程学院, 江苏 南京 210024;
    5. 河南交通职业技术学院, 河南 郑州 450000
  • 收稿日期:2022-06-16 修回日期:2023-04-01 发布日期:2023-05-31
  • 作者简介:牛鹏涛(1985-),男,博士,副教授,主要研究方向为摄影测量与遥感。E-mail:2008051@hnpi.edu.cn
  • 基金资助:
    国家自然科学基金(41671432;41372340)

Automatic road marking extraction method based on vehicle LiDAR point cloud and panoramic image

NIU Pengtao1,2, CAO Yi3,4, ZHANG Enchao5, QI Yang1   

  1. 1. College of Earth Science, Chengdu University of Technology, Chengdu 610059, China;
    2. Henan Polytechnic Institute, Nanyang 473009, China;
    3. Sinopec Northwest China Petroleum Bureau, Urumqi 830011, China;
    4. School of Sciences and Engineering, Hohai University, Nanjing 210024, China;
    5. Henan College of Transportation, Zhengzhou 450000, China
  • Received:2022-06-16 Revised:2023-04-01 Published:2023-05-31

摘要: 针对传统人工方式、人机交互方式提取公路标线成本高、效率低的问题,本文提出了融合车载LiDAR点云和全景影像数据,使用SCGA-Net网络提取并矢量化道路标线的方法,以解决车载LiDAR点云采集过程中因车辆遮挡、道路修补等造成的数据缺失问题。试验表明,该方法道路标线的提取率和正确率均优于仅使用LiDAR点云提取标线的方法,可有效地提升自动驾驶所依赖的高精地图的生产效率。

关键词: LiDAR点云, 全景影像, 道路标线, SCGA-Net网络, 提取

Abstract: In view of the traditional artificial way, the man-machine interactive way to extract road marking cost is high, the problem of low efficiency, put forward integration vehicle LiDAR point clouds and panoramic images as the data source, using SCGA-Net network to extract the road marking method, vector quantization and to solve the vehicle in the process of the LiDAR points clouds gathering caused by vehicle cover, such as road repair data missing problem. Experimental results show that the extraction rate and accuracy of road markers are better than the method of only using LiDAR point cloud to extract road markers, and can effectively improve the production efficiency of high-precision maps that automatic driving relies on.

Key words: LiDAR points clouds, panoramic image, road marking, SCGA-Net network, extraction

中图分类号: