ICA-based multi-temporal multi-spectral remote sensing images change detection

Abstract

Change detection is the process of identifying difference in the scenes of an object or a phenomenon, by observing the same geographic region at different times. Many algorithms have been applied to monitor various environmental changes. Examples of these algorithms are difference image, ratio image, classification comparison, and change vector analysis. In this paper, a change detection approach for multi-temporal multi-spectral remote sensing images, based on Independent Component Analysis (ICA), is proposed. The environmental changes can be detected in reduced second and higher-order dependencies in multi-temporal remote sensing images by ICA algorithm. This can remove the correlation among multi-temporal images without any prior knowledge about change areas. Different kinds of land cover changes are obtained in these independent source images. The experimental results in synthetic and real multi-temporal multi-spectral images show the effectiveness of this change detection approach.