Skip to Main Content
Building boundary is of great significance for building reconstruction and digital mapping. The emergence of airborne LIDAR technology makes rapid extraction of buildings possible. Based on the analysis of existing buildings extraction method, the article proposes an automatic method for extracting 3D building boundaries directly from airborne LIDAR points cloud, which consists of three steps: firstly, building points are segmented by a multi-resolution directional prediction filter method and roof points are then recognized by a series of refinement procedure; Secondly, TIN mode is built by roof points and initial boundaries are extracted by analysis of height differences among vertices and relationships between vertex and edge of TIN facets. Lastly, initial boundaries are regularized by a modified method for roofs with different shapes. Datasets with different roof types are selected to test the proposed approach and results show that it can correctly extract 3D buildings boundaries from discrete LIDAR points cloud.