结合角度纹理信息和Snake方法从LiDAR点云数据中提取道路交叉口

Extraction of road intersection from LiDAR point cloud data based on ATS and Snake

道路交叉口是道路网络组成的重要部分,对道路交叉口的定位与信息提取是掌握道路属性、获取道路网络拓扑结构的重要基础。传统的基于影像的道路交叉口方法对此讨论较少,且存在检测率较低、自动化程度不高等问题,而人工操作成本较高、费时费力。针对从遥感数据中识别与获取道路交叉口信息的问题,提出一种从机载LiDAR点云数据中提取道路交叉口轮廓的方法,该方法的自动化程度较高。首先,利用角度纹理信息(angular texture signature,ATS)对道路交叉口进行粗定位,通过基于点密度在噪声的空间数据库中对所提取的结果进行聚类分析(density-based spatial clustering of application with noise,DBSCAN)和剔除孤立点确定道路交叉口提取区域;然后,引入环形剖面方法进行边缘提取,借助高程信息进行辅助判断,拟合出道路平行边缘;最后,利用Ziplock Snake算法提取出道路交叉口轮廓。实验采用完全自主的研发平台,经历了较多次实验和改进,其结果表明,该方法对道路交叉口的检测率较高,对轮廓提取的精度较高,不仅可行,而且有效。

Road intersection is one of the most important parts of road network, and the positioning and information extraction of the road intersections constitute the basis of road network status monitoring. The traditional image - based research on this problem is so insufficient that there exist such weaknesses as the low detection rate and the need of manipulation by professional workers, which lead to low- level efficiency. In this paper, the authors proposed a method for road intersection extraction from airborne LiDAR point cloud. First, the coarse detection methods based on angular texture signature （ATS）analyzing and density -based spatial clustering with noise （DBSCAN） algorithm were performed to determine the ROI. Then, road edge points were extracted using circular profile in combination with elevation value validating, and the parallel edge lines were fitted. At last, the active Snake function with the ＂Ziplock Snake＂ way energy minimization was used to extract the road intersection contour. The experimental results show that the method presented in this paper has high detection rate and accuracy.