Automatic building extraction from LIDAR digital elevation models and WorldView imagery

This paper proposes an automatic approach for building extraction in suburban and rural areas by integrating the light detection and ranging (LIDAR) data and WorldView imagery. This approach consists of two major steps: building detection and 2D building reconstruction. In the first step, a normalized digital surface model (nDSM) is produced from LIDAR data. Then a building mask is obtained from nDSM based on the elevation and roughness analyses. In the second step, the building mask, along with orthoimage are used to determine preliminary position of the building boundaries. Then a least square line fitting process is applied to produce line segments. Finally the boundaries of buildings are reconstructed by grouping line segments, and intersection of the favorable groups. The proposed approach has shown a 92.5% completeness and 94.75% correctness.