Context-sensitive similarity based supervised image change detection

Abstract

To deal with the problem that traditional satellite remote sensing image change detection methods overestimate changed areas, a context-sensitive similarity based supervised satellite image change detection method was proposed. Both context-sensitive magnitude and direction of change in the vicinity of each pixel by means of local intercept and slope were exploited, and then SVM (support vector machine) with local intercept and slope was used in satellite image change detection. In the experiment for change detection of high resolution bi-temporal multispectral earthquake satellite images including building damage, the results showed that compared to standard SVM, the accuracy of satellite image change detection had been obviously improved, and overestimation of changed areas had been effectively reduced.