1/8 AUTOMATED DAMAGE DETECTION AND VISUALIZATION USING HIGH-RESOLUTION SATELLITE DATA FOR POST-DISASTER ASSESSMENT Masashi MATSUOKA 1 , T. Thuy VU 2 and Fumio YAMAZAKI 3 1 Team Leader, Earthquake Disaster Mitigation Research Center, NIED, Kobe, Japan. 2 Researcher, Earthquake Disaster Mitigation Research Center, NIED, Kobe, Japan. 3 Professor, Faculty of Engineering, Chiba University, Chiba, Japan. SUMMARY The focus of this study is to thoroughly exploit the capability of very high-resolution (VHR) satellite imagery such as IKONOS and QuickBird for disaster mitigation. An efficient automated methodology that detects damage is implemented to derive the rich information available from VHR satellite imagery. Consequently, the detected results and the VHR satellite imagery are attractively presented through a fly-over animation and visualization. The aim is to assist the field-based damage estimation and to strengthen public awareness. The available IKONOS and QuickBird data captured after the Bam, Iran earthquake in December, 2003 was employed to demonstrate the competence of our automated detection algorithm and fly-over animation/visualization. Our results are consistent with the field-based damage detected results. INTRODUCTION For decades, remote sensing techniques have been important in grasping damage information caused by earthquakes. Medium resolution satellite data like SPOT, Landsat [1, 2] or ERS [3] is mainly used to identify the extent of the damage. Damaged buildings can be detected using aerial photographs [4]. Recently, very high-resolution (VHR) imagery from commercial satellites such as IKONOS and QuickBird, which can be rapidly acquired, is becoming more powerful and is providing information on natural and/or man-made disasters in the early stages. Both visual interpretation and automated analysis are currently used to detect damaged buildings, but the latter has yet to be reliably implemented. The conventional method for detecting damage caused by an earthquake is to compare pre- and post-event images. This approach has also been developed for VHR data. For example, a new overlay method between pre- and post-event images was based on artificial neural networks [5]. However, it is unrealistic to obtain images of the stricken areas before a disaster, and archived data with clear and suitable images is limited. Therefore, we have been studying an automated detection method using only post-event images in order to efficiently use the instantaneous acquisition ability of helicopters and airplanes [6, 7]. This study follows our previous work by investigating the suitability of this edge-based technique [8, 9], which was developed by examining the relationship between aerial images and detailed damage survey data obtained after the 1995 Kobe earthquake, on VHR satellite imagery.