Abstract: Rapid and accurate extraction of rural homestead area is of great significance for urban and rural planning, housing and land integration, and beautiful rural construction. In this paper, one village in Shaanxi province is taken as the research area. Based on UAV high-resolution remote sensing data, multi-scale segmentation and supervised classification are used to extract the building contour. The number of building floors is then estimated based on oblique photogrammetry. Finally, a fast estimation of homestead area is carried out. The overall accuracy of building extraction in the study area is 95.8%, and Kappa coefficient is 0.892 5. The total area of residential land in the study area is 36 417 m². The results show that this method can accurately and effectively extract the homestead area, and provide a reference for UAV remote sensing in the extraction and identification of buildings and the rapid determination of homestead area.

Key words: UAV, remote sensing, homestead area, Western China, rural area