AN AUTOMATIC APPROACH FOR CHANGE DETECTION IN LARGE-SCALE REMOTE SENSING IMAGES

In this paper, we present an automatic approach for change detection in a large and complex image scenario. The proposed technique takes advantages of automatic registration algorithm and change detection method that jointly measures the spatial invariant but spectral variant features in the considered bitemporal remote sensing image pair. Two classes of pseudo training samples, which associated to the change and no-change two classes, are automatically generated by analyzing the change representation information from both global and local perspectives. Finally, the robust classifier, i.e., linear support vector machine (LSVM), is used to identify the binary changes in the whole image scenario using the pseudo training samples. Experimental results obtained on a pair of real bitemporal Landsat-8 OLI images covering a large scene confirmed the effectiveness of the proposed method.