欢迎访问中国科学院大学学报,今天是

中国科学院大学学报 ›› 2023, Vol. 40 ›› Issue (6): 800-809.DOI: 10.7523/j.ucas.2022.024

• 电子信息与计算机科学 • 上一篇    下一篇

基于点密度与卡尔曼滤波的路面标识线提取方法

李晓宇1,2, 周梅1, 王金虎1, 姚强强1   

  1. 1. 中国科学院空天信息创新研究院 中国科学院定量遥感信息技术重点实验室, 北京 100094;
    2. 中国科学院大学光电学院, 北京 100049
  • 收稿日期:2022-02-17 修回日期:2022-03-28 发布日期:2022-04-07
  • 通讯作者: 周梅,E-mail:zhoumei@aoe.ac.cn
  • 基金资助:
    国家自然科学基金(61771456)资助

Road marking detection and extraction method based on neighborhood density and Kalman filter

LI Xiaoyu1,2, ZHOU Mei1, WANG Jinhu1, YAO Qiangqiang1   

  1. 1. CAS Key Laboratory of Quantitative Remote Sensing Information Technology, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100094, China;
    2. School of Optoelectronics, University of Chinese Academy of Sciences, Beijing 100049, China
  • Received:2022-02-17 Revised:2022-03-28 Published:2022-04-07

摘要: 提出一种基于邻域点密度和卡尔曼滤波的车载激光点云路面标识线提取与补全方法。该方法由3个部分组成:1)利用时间信息将路面点云数据分割为单条扫描线;2)根据邻域点密度自动生成卷积核以提取标识线轮廓点;3)结合最小二乘算法拟合标识线轮廓,并利用卡尔曼滤波算法填补缺损标识线。经实验验证,最终标识线提取结果的平均中心点位置偏差为0.04 m,平均方向偏差为0.04°,平均完整度为99.69%。该方法降低了点云密度分布不均对路面标识线检测精度的影响,可有效提高标识线检测的完整度。

关键词: 车载激光扫描, 点云, 卷积, 路面标识线, 卡尔曼滤波

Abstract: This paper presents a methodology for detection and extraction of road marking based on neighborhood point density and Kalman filter from mobile laser scanning in three steps:1) Segmenting road point clouds into scan lines; 2) Generating convolution kernel based on point density in neighborhood for extracting road marking contour points; 3) Fitting contour line using least squares algorithm and completing omitted road markings using Kalman filter. A quantitative validation shows that the average deviation of center points and orientation are 0.04 m and 0.04°, respectively, and the average completeness of results are 99.69%. The proposed method reduces the influence of unevenly distributed points on road marking extraction and improves the overall extraction completeness.

Key words: mobile laser scanning, point clouds, convolution, road markings, Kalman filter

中图分类号: