Building extraction from VHR multi-spectral images using rule-based object-oriented method: A case study

摘要

Object-oriented classification has been demonstrated a promising method for large-scale detailed urban structure mapping using very high-resolution space-borne or airborne remote sensing images. In the paper, the object-oriented classification method for building extraction using pan-sharpened IKONOS multi-spectral images was applied to roof mapping combined with Lidar data. The scheme to produce the vector polygon database of buildings includes the following steps: (1) image pre-processing and derivation of secondary inputs to image segmentation and classification procedures; (2) segmentation of processed data layers into image objects; (3) classification of image objects; (4) export classified map objects and average building height data; and (5) polygon generalization and/or geometrical regularization. The experimental result is visually satisfactory and may suit overall investigations of building development.