An automatic building extraction method from high resolution satellite image

Building extraction plays an important role in GIS, image understanding and military reconnaissance. However, the existing methods cannot guarantee the accuracy when run automatically. Thus, by analyzing the characters of building and non-building regions in satellite images, a novel automatic and time-efficient building boundary extraction method, with the aid of corresponding ordinary map, is proposed. Firstly, the candidate edges of buildings are obtained by line segment analysis. Then, the candidate building boundaries are determined by region segmentation, analysis and region merging successively. Finally, all candidate building boundaries are checked and verified according to the prior geometric knowledge about buildings. The proposed method was implemented and tested on the Google satellite maps. The physical experiments demonstrated the accuracy and efficiency of this method.