Change detection for reconstruction monitoring using very high resolution optical data

Abstract

This thesis studies change detection for reconstruction monitoring using VHR optical data. Several pixel-based methods are compared and evaluated for reconstruction monitoring of urban areas, which are classical image differencing, iteratively reweighted multivariate alteration detection (IR-MAD), IR-MAD incorporating textural information and direct multidate classification. In this study, the performances of different agorithms are evaluated quantitatively in terms of various accuracy measures. Compared with manually acquired reference data, the accuracy of the change detection map extracted by IR-MAD combined with textural information is satisfying on the whole.