Automated extraction of 3D vector topographic feature line from terrain point cloud

Abstract

This paper presents an automated topographic feature lines detection method that directly extracts 3D vector topographic feature lines from terrain point cloud. First, signed surface variation (SSV) is introduced to extract the potential feature points. Secondly, the potential feature points are segmented to different clusters by combining region growing segmentation and conditional Euclidean clustering. In order to extract feature points, the potential feature points in each cluster are iteratively thinned using a HC-Laplacian smoothing method with SSV weighted taken into account. Besides, SSV-based and elevation-based simple rules are added for accelerating this thinning process. Finally, the feature lines are obtained by constructing the minimum spanning tree of the extracted feature points. By comparing with manually digitized reference lines, the correctness and the completeness of extracted results are about 80% or even higher, which are much higher than those extracted by D8 algorithm.

Keywords:

• Feature line extraction
• point cloud
• signed surface variation
• HC-Laplacian smoothing
• minimum spanning tree