Abstract:

Aiming at the noise interference in detecting road centerlines from airborne LiDAR data, with multi-feature road saliency, a road centerline extraction method based on the minimum scale factor determined by the maximum road width to cast tensor voting algorithm is proposed. The pre-processed three-dimensional road point clouds are transformed into a two-dimensional intensity image. Ball voting is executed with the minimum scale factor to eliminate the road edge parts with polarity. To enhance the salient linear features, ball and stick tensor votings are carried out on the road image to fill the gaps and smooth the boundary with the recalculated minimum scale factor. The accurate road centerlines are obtained by thinning algorithm. Compared with the mathematical morphology method, tensor voting algorithm can extract road centerlines with higher accuracy from the noisy image.

Key words: tensor voting, airborne LiDAR data, road centerlines, the minimum scale factor, multi-feature saliency

摘要：

针对机载LiDAR数据中道路骨架线检测存在的噪声干扰问题，结合道路多层特征显著性，提出了一种基于道路最大宽度快速确定最小尺度因子的张量投票道路骨架线提取方法。将预处理后三维道路点云转化成二维强度图像，最小尺度因子参与图像球张量投票，利用极性特征分割道路边缘点；为了进一步增强道路线状特征，利用新的最小尺度因子再次进行球张量投票和棒张量投票，填补道路空洞，顺滑道路边界；细化处理获取道路骨架线。与数学形态学方法相比，该方法在噪声背景的道路数据中提取的道路线精度更高。

关键词: 张量投票, 机载LiDAR点云, 道路骨架线, 最小尺度因子, 多层特征显著性