Polcz, Péter and Benedek, Csaba (2014) 3D mesh generation from aerial LiDAR point cloud data. In: Seventh Hungarian Conference on Computer Graphics and Geometry,, 2014, Budapest.

Preview

Text Polcz_Grafgeo14.pdf Download (3MB) | Preview

Abstract

Three dimensional urban scene modelling became important issue in the last few years. Beside visual experience, 3D city modelling has gained a significant function in diverse analysing tasks, however the amount of data requires a high level of automation of model generation. In this work, we introduce an automatic and robust algorithm which produces detailed 3D virtual city models by analysing high resolution airborne LiDAR point clouds. Using the idea of the surface normal based roof segmentation we have designed a procedure, which takes into account the boundaries of each roof segment, so that the adjacent segments connect without gaps. We have developed an algorithm to detect 3D edge lines of the rooftops. Since the applied triangulation methods operate on the whole convex hull of the input points, hollow outer parts of the roof segments are filled in with false triangles. To solve this problem, we have proposed a method using a Markov Random Field, in which we filter out the incorrect triangles lying on the concave parts.

Update Item