Abstract: Automatic recognition and classification of building seismic disaster information is the key step of the seismic disaster evaluation. And its precision directly affects the result of loss evaluation. With the development of high-resolution remote sensing image, the traditional pixel-based classification is not able to meet the requirement. In order to fully excavate the texture, shape and relationship information of remote sensing data, we introduced an object-oriented method to improve the classification precision. Aiming at building seismic damage detection, object-oriented method was used in this paper to classify the post -earthquake aerial remote sensing images of the 2008 Wenchuan M8.0 earthquake. Compared to pixel-based classification, object-oriented method significantly improved the effect of damage detection.

Key words: Object-oriented classification, Damage detection, High-resolution remote sensing image, Wenchuan M8.0 earthquake