Automatic identification of watercourses in flat and engineered landscapes by computing the skeleton of a LiDAR point cloud. (September 2017)

• Record Type: Journal Article
• Authors: Broersen, Tom; Peters, Ravi; Ledoux, Hugo
• Is Part Of: Computers & geosciences. Volume 106(2017)
• Page Start: 171
• Page End: 180
• Publication Date: 2017-09
• Subjects: LiDAR -- Skeleton -- Watercourse -- Medial Axis Transform; Environmental policy -- Periodicals; 550.5
• DOI: 10.1016/j.cageo.2017.06.003
• Languages: English
• ISSNs: 0098-3004

Abstract

Abstract: Drainage networks play a crucial role in protecting land against floods. It is therefore important to have an accurate map of the watercourses that form the drainage network. Previous work on the automatic identification of watercourses was typically based on grids, focused on natural landscapes, and used mostly the slope and curvature of the terrain. We focus in this paper on areas that are characterised by low-lying, flat, and engineered landscapes; these are characteristic to the Netherlands for instance. We propose a new methodology to identify watercourses automatically from elevation data, it uses solely a raw classified LiDAR point cloud as input. We show that by computing twice a skeleton of the point cloud—once in 2D and once in 3D—and that by using the properties of the skeletons we can identify most of the watercourses. We have implemented our methodology and tested it for three different soil types around Utrecht, the Netherlands. We were able to detect 98% of the watercourses for one soil type, and around 75% for the worst case, when we compared to a reference dataset that was obtained semi-automatically. Highlights: We describe a 2D and a novel 3D skeleton method for the automatic identification of watercourses. We describe a generic methodology to merge centrelines from different methods. We achieve detection rates of up to 98% when compared to out reference dataset.