Abstract: Many studies have focused on the extraction of buildings from Light Detection and Ranging (LiDAR) point cloud information, but few have investigated the process in rural areas where vegetation and buildings are of similar heights and interconnected. With Sihushan County in Hunan Province which has typical rural building characteristics as study area, the building information is extracted using LiDAR point cloud data. A modified morphological filter in which gradient and gradient direction of primitive point were used to constraint the area of filtering is adopted. The interpolated ground points and primitive points were used to obtain a digital elevation model (DEM) and a digital surface model (DSM), and the two were subtracted to derive a normalized DSM (NDSM). Then, a transformation of the sign watershed was conducted under control of both height and gradient to obtain ground objects. Finally, using built feature indicators, building objects were identified based on a maximum likelihood classification. Results show that the user accuracy and producer accuracy of the building extraction are both greater than 90%, and the Kappa coefficient is greater than 0.8, which suggest that the proposed method achieved good results in building extraction in rural areas.

Key words: LiDAR, morphological filter, watershed algorithm, features, maximum likelihood classification