Building Damage Detection of the 2003 Bam, Iran Earthquake Using Quickbird Images Based on Object-based Classification

In resent years, remote sensing technologies are utilized in disaster management as a means of information gathering in a large scale disaster. QuickBird captured a clear image of Bam City, eight days after the 2003 Bam, Iran, earthquake. The city was also observed by QuickBird about three months before the event. In this paper, we perform building damage detection based on land cover classification. Two supervised classification methods are employed; one is pixel-based classification and another object-based classification. First, these two methods are applied to building detection using the pre-event image. A reasonable result is obtained for the object-based classification while salt-and-pepper noises are observed for pixel-based classification. The object-based approach is further applied to the post-event QuickBird image to detect debris areas. Comparing the result by object-based classification with that from visual inspection, a reasonable level of accuracy is obtained for debris locations although further improvements are suggested.