基于车载激光扫描技术的道路标线自动提取

Automatic extraction of road markings based on vehicle borne laser scanning

为了提高基于车载点云数据的道路标线提取精度,本文以车载点云数据为数据源,对点云数据进行转换,生成强度特征图,并利用深度学习算法和KD树聚类分割算法实现道路标线的提取和矢量化。试验结果表明,利用该算法可以有效提取道路标线数据,准确反映道路信息。其中,提取的人行横道线精度达到93.57%,道路标线均交并比达到86.32%。此外,该算法还能在实现道路标线自动提取的基础上对道路标线数据进行分类,为城市道路标线的快速、准确提取提供思路。

In order to improve the accuracy of road marking extraction based on vehicle borne point cloud data,vehicle borne point cloud data is used as the data source,and it is transformed to generate intensity feature maps,and then deep learning algorithms and KD tree clustering segmentation algorithms are used to extract and vectorize the road markings.The experimental results show that road marking data can be effectively extracted and road information can be accurately reflected by using these algorithms.Among them,the accuracy of the extracted zebra crossing is 93.57%,and the average intersection to intersection ratio of the extracted road markings is 86.32%.Furthermore,this algorithm can also classify road marking data based on the road markings from automatic extraction,so it will provide ideas for rapid and accurate extraction of urban road markings.