สุพรรณี ตะทะนอง. Object Based Change Detection after a Disaster : The Tsunami Case. Master's Degree(Computer Engineering). King Mongkut's University of Technology Thonburi. KMUTT Library.. : King Mongkut's University of Technology Thonburi, 2552-10-25.
Object Based Change Detection after a Disaster : The Tsunami Case
Abstract:
The tsunami of December 26, 2004 was one of the worst disasters in recent human history. The event caused destruction in natural resources and settlements. Following a disaster, change detection is a prereqllisite for quick assessment of the damage. To assess the severity of devastation, most damage assessments focus on the destruction of nlan-made objects, particularly buildings. To examille changes in buildings after a disaster, building detection is the most important task. Despite tIle rich literature on building extraction, existing approaches all have weaknesses. Most approaches misdetect small-sized bllildings in residential areas. As a result, response and rehabilitation activity following a disaster may not be planned accurately. The other major concern is that the pre-disaster and post-disaster images may not be captured at tIle same resolution or may even be acquired from different sensors. In this research, we attempt to find the solutions that address these concerns. To overcome the first problem, we develop a robust rectangular building extraction that can detect both small-sized buildings in residential areas and large-sized buildings in illdustrial areas. In our work,
we use both edge detection and region growing approaches to supplement each other. To find a solution for the second concern, we develop a change detection method by applying an object oriented technique in conjunction with an intelligent agent technique. In our experiments, our rectangular building detection could identify buildings with 75 percent accuracy. Overall, with accurate input from the building extraction, our change detection can report real changes in area, intensity and structural information including dimension and orientation. With our promising results, in conjunction with fllture enhancements, our approach can provide helpful information about the state of bllildings after a disaster. As a result, rehabilitation activity can be provided accordingly.