An Improved Approach of Information Extraction for Earthquake-Damaged Buildings Using High-Resolution Imagery

The development of remote sensing technology, especially the availability of high-resolution satellite imagery, has been applied to building recognition, hazard investigation and rapid pre-evaluation in post-earthquake management. Existing pixel-oriented approaches which are commonly used for satellite high-resolution imagery have limitations in information extraction, ground object classification, and processing speed. This paper presents an object-oriented method to extract earthquake-damaged building information using high-resolution remote sensing imagery of the 5.12 Wenchuan Earthquake. This method segmented the whole image into non-intersecting pieces of image objects, and then classified these pieces to extract damaged/undamaged buildings using image features such as spectral characters, textures, shapes, and their contexts. The results show a higher-precision classification than conventional methods.