Spectral and spatial classification of earthquake images by support vector selection and adaptation

The aim of this study is to extract homogenous and edge regions from a post-earthquake Quickbird satellite image with high resolution and to combine this spatial information with spectral information in classification of earthquake damage. In order to extract the homogenous and edge regions from the image, a spatial filtering approach and Canny filter were used. A novel method called support vector selection and adaptation (SVSA) was used in classification of earthquake damage. Pixel and texture-based classification were separately carried out in order to show their comparative classification performance. For implementation, a small region from city of Bam in Iran was selected.