Automatic building extraction from airborne LiDAR point cloud based on mean shift segmentation

Abstract

Building extraction is an important part for smart city construction. This paper proposes a novel method for automatic building extraction from airborne LiDAR point cloud. In the present study, filtering was first applied to point cloud, which could help obtain elevated points for generating the DTM. The building-candidate points were then obtained by setting a threshold from the DTM. To distinguish the tree points from building points, three constraints, namely, area constraint, point density constraint and root mean square error constraint were applied to the building-candidate points. By comparing with the reference data generated manually, the evaluation result shows that the proposed method could yield a good performance.