Building boundaries extraction from points cloud using an image edge detection method

Building digitalization is an important component of digital city development. Terrestrial laser scanning (TLS) technology provides a high-precision and high-density data source to guarantee building digitalization and reconstruction. However, when these large scattered points cloud are used to build a 3D digital model, the points of building boundaries extraction is an essential issue. In this paper, the edge detection method for 2D images is extended to the 3D points cloud to achieve rapid, complete and accurate boundaries points for building. Firstly, based on the coplanar conditions, the points cloud data of a building are divided into different patches and then converted into 2D images according to the depth dimension of each patch. Secondly, the points of each plane are then extracted by an improved Laplace image edge detection method. Finally, the boundary points extracted by the method proposed are validated using tested data. It shows that the proposed method is effective and robust: the extraction accuracy for the building boundary points is up to 85%, notably for cases with more complicate features.