Two dimensional change detection methods for satellite images

Change detection using remotely sensed images has many applications such as urbanmonitoring, land-cover change analysis, and disaster management. Therefore, weinvestigated two dimensional change detection methods in this thesis study. We grouped theexisting methods in the literature into four categories as: pixel based, transformation based,texture analysis based, and structure based. Besides testing the existing ones, we introducefour new change detection methods as: fuzzy logic based, shadow detection based, localfeature based, and bipartite graph matching based. The last two novel methods are thebasis for structural analysis of change detection. We compare three thresholding algorithmsand measure their effects on the performance of change detection methods. In testing theexisting and newly introduced change detection methods, we use a total of 35 panchromaticand multi-spectral Ikonos image sets. We provide quantitative test results as well as theirinterpretation.