ABSTRACT
The last three decades have seen a rapid increase in the number of natural disasters, especially from strong winds such as hurricanes and tornadoes. Although natural disasters are inevitable, mitigation at the right time in the right way, by providing quick and accurate information, will prevent a natural event from ending up as a catastrophe. Access to damaged areas by field investigation is time-consuming and many a time difficult, immediately after a major disaster. However, damage investigation in a bird’s eye perspective can ease the difficulty and enable a quick survey. Identification of damage from remote-sensing images has enabled a breakthrough in the current era, thereby speeding up damage investigations. The present work introduces a novel technology for detecting strong-wind-disaster areas and estimating the damages from post-disaster aerial/satellite imageries. Initially, the wind-disaster area is identified from a low-resolution remote-sensing post-storm imagery by using a newly developed technique called texture-wavelet analysis. Once the damaged area is identified, damage estimation is performed by identifying the damaged buildings in the disaster area and then estimating the percentage area of roof damage. For performing damage estimation, high- resolution remote-sensing post-storm imagery of the same location is used. In the present work, the image segments of damaged buildings are distinguished from the image segments of non-damaged buildings, by a wavelet-based feature extraction, artificial neural network, and Support Vector Machine classification methods. The percentage area of the damaged roof is determined by performing texture-wavelet analysis on the image segments of damaged buildings. A comparison of the present method with the conventional change-detection method is also done on the same imagery. Results are validated from visual interpretation data and field investigation data.
Key words: Remote-sensing imageries; texture-wavelet analysis; artificial neural network; Support Vector Machine; strong-wind disaster
