Image fluctuation model for damage detection using middle‐resolution satellite imagery

A damage detection method using middle-resolution satellite images, the Image Fluctuation Model method, is proposed, which employs a stochastic model of the digital number (DN) fluctuation in a normal condition and its significance test. The DN fluctuation model is formulated by considering an imaging process of a satellite sensor and an image registration process. A resulting thematic map is created based on a confidence level (1-significance level), which is defined on a pixel-by-pixel basis as follows: the minimum significance level at which the null hypothesis, that the pixel DN can be considered as a sample of the DN fluctuation model, is rejected. The confidence level provides the model-based probability of ground surface change. The method is applied to the 2003 Bam, Iran earthquake using images acquired by Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) of Terra. The receiver operating characteristic curves of the method showed better detection performance than temporal image differencing or temporal image ratioing. Though the detection performance of building damage was not comparable to visual inspection on a building basis using high-resolution images of QuickBird, the confidence level map shows similarity at the district level to damage assessment results using high-resolution images.