A Three-class Change Detection Method For SAR Images Based On Evolutionary Multiobjective Optimization

Change detection in remote sensing images involves the analysis of two registered images acquired over the same geographical area at different times in order to identify changes that may have occurred in the study area between the two times considered. In the last decades, it has attracted widespread interest due to a large number of applications in diverse disciplines, such as investigations of forest resources, studies on land use/land cover dynamic detection, assessments of environment disaster, arrangements of urban growth, and monitoring of national defense, etc.This paper presents a three-class change detection method for SAR images based on evolutionary multi-objective optimization. This method is able to identify the changes occurred in multi-temporal SAR images, which is different to the traditional bi-class change detection methods. The proposed approach further classifies the changed regions into positive and negative changes. Therefore, the proposed technique can improve the efficiency and effectiveness of the traditional change detection methods. We have finished two main tasks as following:(1) We generate the difference image by improving the traditional log-ratio and method ratio operators. Second, we analyse the difference images by using classification methods.(2) This paper proposes a three-class change detection method based on evolutionary multi-objective optimization. In the procedure of classifying the three-class difference images, we incorporate local information into the fuzzy clustering to reduce the effect of speckle noise. And this paper proposes a framework for three-class change detection and a novel method for generating three-class difference images.Finally, experimental results on synthetic and real SAR data sets show the effectiveness of the proposed method.