A building extraction method using shadow in high resolution multispectral images

Abstract

Buildings, roads, and squares are all man-made objects which have similar spectral feature. The similarity leads high false alarm rate of building extraction methods that only depend on spectral feature. In high resolution multispectral images, the shadow of buildings are visible, thus, based on the relations of buildings and shadow, this paper proposes a shadow detection method, extracts building sample points exactly depending on shadow using the similarity of angle between shade line and shadow line in one image, then perfectly achieves building extraction. The method reduces the difficulty of classifying buildings from other similar man-made objects, moreover, the detection accuracy is greatly improved.