Multi-scale object-oriented building extraction method of Tai'an city from high resolution image

Based on high-resolution remote sensing images and the eCognition Developer platform, the research described in this paper makes full use of the rich spectral, spatial, textural and feature geometry information contained in high-resolution images acquired from Quickbird. The object-oriented, multiscale segmentation method as well as the nearest neighborhood and membership function classification methods are applied to classify the experimental area into five land categories; i.e. residential building, green space, road, leisure area and bare area, As a result, the residential building information is extracted eventually. The experiment shows that compared to the conventional pixel-by-pixel classification method, the object-oriented classification method can effectively avoid the fragmentation of the segmented regions and provide a more complete, accurate and efficient method of land extraction.