Detecting shadows in high-resolution remote-sensing images of urban areas using spectral and spatial features

Abstract

In accordance with the characteristics of urban high-resolution (HR) remote-sensing images, we propose a shadow detection algorithm that combines spectral and spatial features. Rather than pixel-based shadow features, the proposed features are based on shadow regions obtained by the object-based segmentation method. First, based on the shadow ratio map, the candidate shadow pixels are acquired by the Otsu method. The candidate shadow regions can be identified using connected component analysis. In the candidate shadow regions, shadow spectral and spatial features are calculated. With these two features, the true shadow regions can be distinguished from candidate shadow regions. Experiments and comparisons indicate that our proposed algorithm is feasible and effective for shadow detection in both aerial and satellite images.