Building Height Extraction from GF-7 Satellite Images Based on Roof Contour Constrained Stereo Matching

Building height is one of the basic geographic information for planning and analysis in urban construction. It still very challenging to estimate accurate complex buildings from satellite images, especially with podium. This paper proposes a solution building estimation GF-7 images by using roof contour constrained stereo matching algorithm DSM (Digital Surface Model) based bottom elevation estimation. First, an object-oriented proposed on extract image, generated then used obtain elevation. Second, conducted between backward forward image blocks, which difference standard deviation maps similarity measure. To deal multi-height problem podium buildings, gray adopted segment re-matching find out their actual heights. Third, obtained through top bottom, evaluation calculated according histogram statistics buffer DSM. Finally, two collected Yingde, Guangzhou, Xi’an, Shanxi, are performance evaluation. Besides, aerial LiDAR point cloud absolute accuracy The results demonstrate that compared other methods, our obviously improves high-rise buildings. MAE (Mean Absolute Error) estimated heights Yingde 2.31 m, approximately 1.57 m 1.91 respectively. Then RMSE (Root Mean Square 2.01 2.57 m. As Xi’an dataset 7 40 1.69 2.34 method can be effective extraction images.