Classification and extraction of road markings based on mobile LiDAR point cloud data

Journal Information

J. Hefei Univ. Tech. (Nat. Sci.), Jun.2024, Vol.47 No.6: 843-848

DOI: 10.3969/j.issn.1003-5060.2024.06.019

Authors

GAO Fei $ ^{1} $, WU Yan $ ^{' } $an $ ^{2} $, XIAO Xinfeng $ ^{2} $, YUAN Bin $ ^{2} $, ZHANG Shufeng $ ^{2} $, XIE Ronghui $ ^{2} $

(1. School of Civil and Hydraulic Engineering, Hefei University of Technology, Hefei 230009, China; 2. Anhui Kaiyuan Highway and Bridge Co., Ltd., Hefei 230093, China)

Abstract and Keywords

Abstract: It is difficult to extract road markings by using mobile LiDAR point cloud data. To solve this problem, this paper adopts an improved road marking classification and extraction method based on point cloud feature image. Firstly, the road point cloud is projected to generate the point cloud feature image. Through the image gradient analysis, image binarization and connected domain analysis, the road marking pixels are further extracted. Then, after back projection to 3D point cloud, Gaussian mixture model is used to finely optimize the road markings, so as to extract the complete road marking point cloud. Finally, the road marking point cloud is classified and extracted through template matching classification. The experimental results show that the accuracy, integrity and comprehensive evaluation of road markings extracted by this method in different road environments exceed 90%.

Keywords: road markings; point cloud feature image; image gradient analysis; Gaussian mixture model; template matching classification

作者信息

高飞 $ ^{1} $，吴言安 $ ^{2} $，肖信峰 $ ^{2} $，袁斌 $ ^{2} $，张树峰 $ ^{2} $，谢荣晖 $ ^{2} $

(1. 合肥工业大学土木与水利工程学院, 安徽合肥 230009; 2. 安徽开源路桥有限责任公司, 安徽合肥 230093)

摘要和关键词

摘要: 利用车载激光道路点云提取道路标线的难度较大。针对此问题, 文章采用一种改进的基于点云特征图像的道路标线分类提取方法。首先将道路点云投影生成点云特征图像, 通过结合图像梯度分析、图像二值化和连通域分析等操作, 进一步进行道路标线像素提取; 然后反投影到三维点云后, 利用高斯混合模型对道路标线精细优化, 从而提取出完整的道路标线点云; 最后通过模板匹配分类, 对道路标线点云进行分类提取。实验结果表明, 该方法对不同道路环境下道路标线提取的准确度、完整度以及综合评价都超过 90%。

关键词: 道路标线；点云特征图像；图像梯度分析；高斯混合模型；模板匹配分类

Funding

国家自然科学基金资助项目（41904010）；安徽省自然科学基金资助项目（2008085MD115）

合肥工业大学学报(自然科学版)

Navigation