Land change detection using multivariate alteration detection and Chi squared test thresholding

In this paper we shall describe a statistical approach for land change detection based on multivariate alteration detection (MAD) transformation combined with a thresholding method based on Chi squared test. Unlike the most other multivariate change detection techniques, the MAD analysis is invariant to linear and affine transformations of the input data. Consequently, it is insensitive to linear differences in atmospheric conditions or sensor calibrations of multitemporal acquisitions. Detected change objects by the MAD variates are then extracted by means of the studied thresholding technique. We proposed also post-processing of the change detected using the MAD variates by means of maximum autocorrelation factor (MAF) analysis. A case study with SPOT-HRV multispectral data before and after a flood event occurred in November 2000 shows the usefulness of the proposed MAD/Chi-2 and MAF/MAD/Chi-2 change detection schemes according to the ground truth of the study zone.