Façade Separation in Ground-Based LiDAR Point Clouds Based on Edges and Windows

Abstract

Extraction of individual facades from building point clouds collected by ground-based LiDAR is vital for urban mapping and modeling. It is a challenging task due to the complexity of facades, and very few studies have been conducted. In this paper, we present a new facade separation method that can divide connected facades into building instances using point coordinates only. The proposed method consists of two steps. The first step is extracting edges and windows from building point clouds. In this step, an improved window detection method which can detect complex windows, such as bay windows, is proposed. In the second step, the separation problem is solved by minimizing a new objective function, which considers both the edge intersections and wall elevations. This objective function is constrained by the number and positions of windows. After the optimization, a subset of individual facades defined by dividing lines will be selected from potential facade candidates. The proposed method is tested in two ground-based LiDAR datasets, which contain facades of various building architectures. One dataset is acquired by static terrestrial laser scanning, and mobile LiDAR collects the other dataset. The overall precision and recall of our facade separation results are 85.5% and 74.6%, respectively. We also compare our methods with existing window detection approaches and other possible facade separation algorithms. These experiments demonstrate the effectiveness and advantages of the proposed methods.