Accuracy Improvement of Building Damage Detection Using High-Resolution SAR Images Observed from Different Directions

Abstract

The effect of observation direction on building damage detection from high-resolution SAR images is discussed by using two pairs of TerraSAR-X images and building damage data in the area damaged by the 2010 Haiti earthquake in order to improve the detection accuracy. The ascending pair and the descending pair of the images observed before and after the earthquake are analyzed. Linear discrimination analysis is applied to the pairs of the images in order to identify the collapsed buildings. The result shows that the accuracy of the damage detection depends on the relation between the observation direction and the building direction. The analysis based on the two pairs of the images achieves higher accuracy than the analysis based on a pair of the images.