Automatic change detection in remote sensing images using level set method with neighborhood constraints

Abstract

An automatic change detection (CD) method based on level set evolution in remote sensing images is proposed. The CD problem is formulated as a segmentation issue to discriminate the changed class from the unchanged class in the difference images. The strategy of the level set initialization is considered and neighborhood constraints are added to the level set energy model. In addition, a coarse-to-fine procedure is adopted. A chief advantage of our approach is to be able to obtain correct results even when the difference image contains different types of changes. Furthermore, the proposed method is robust against noise and yields smooth boundaries of changed regions without manual parameter adjustment. We implement the proposed method in a multiresolution framework and validate the algorithm systematically with a variety of remote sensing images by low- as well as high-spatial resolution sensors, including Landsat-5 TM, SPOT5, IKONOS, etc.