Cyclone damage detection on building structures from pre- and post-satellite images using wavelet based pattern recognition

The majority of building structural losses during the past few decades has been due to wind induced damage, especially tropical cyclone damages. Rapid identification of damage locations as well as damaged buildings, and appropriate maintenance, can diminish the impact of such natural disasters. This paper describes detection of damage to cyclone-prone building roof structures from pre- and post-storm satellite images of the shores of Punta Gorda before and after the Hurricane ‘Charley’ 2004 disaster using a wavelet-based change detection method. Damaged buildings are automatically identified using wavelet-extracted statistical features and by edge detection, and classification using Artificial Neural Network (ANN) and Support Vector Machine (SVM). A comparison analysis is then carried out by comparing these results with results obtained using conventional change detection methods. It is observed that the wavelet-based change detection method yields identification information of damaged buildings superior to that obtained using conventional methods. In this work, the percentage of the damaged area of each damaged building is also calculated by a newly introduced texture-wavelet analysis on roof-tops, and the results are validated by counting the damage pixels manually. A positive increase in the extracted statistical features is observed as the percentage area of damage increases, which adds to the accuracy of the identification method.