Building Damage Detection of the 2004 Nagapattinam, India, Tsunami Using the Texture and Spectral Features from IKONOS Images

In this paper, we implemented and evaluated the proposed method for building detection and classification using high-resolution imagery (IKONOS) data. Nagapattinam province of Tamil Nadu, which was strongly attacked by the 2004 Indian Ocean tsunami, is selected as the demonstration site. IKONOS sets acquired on June 10, 2003, (pre-tsunami)–December 29, 2004, (post-tsunami) are used in this study. This approach follows the standard scheme of image analysis: building detection, feature extraction and selection, and classification. At the first level, we studied the damaged buildings from imagery data set. In the second level, 28 texture features and 3 spectral features are extracted, and then 10 best features are selected for next stage by using the statistical dependency feature selection method. In the classification, the proposed method feature mean ratio is used to classify the damaged buildings. The proposed method has been applied on an IKONOS imagery in Nagapattinam city, and visual validation demonstrates that the proposed method provides promising results. The results proved that the proposed work has produced a higher accuracy assessment (96.4%) when compared to earlier works.