Change Detection In SAR Images Using Multiobjective Clustering With Nonlocal Means

In recent years, with the development of remote sensing technology, synthetic aperture radar attracts more and more people’s attention because it is not affected by illumination, especially in the field of change detection. Change detection means that the computer detect the difference between the two images of the same place at different times. This thesis conducts a study about change detection in synthetic aperture radar images and presents a method of change detection in synthetic aperture radar images based on multiobjective clustering with nonlocal means. The main research results consists of two aspects:1. The difference map analysis methods based on multiobjective clustering is put forward. The image details and noise present in an image at the same time, thus the trade-off between detail preserving and noise insensitiveness plays a critical role in change detection tasks. We define two objective functions which represent detail preserving and noise removing respectively. Accordingly, the change detection problem is transformed into a multiobjective optimization problem and optimized using a multiobjective evolutionary algorithm. As a result, we obtain a series of Pareto optimal solutions with different degree of detail preserving and noise removing, so users can choose a suitable solution which is an appropriate answer to their specific problem.2. The change detection method based on multiobjective clustering with nonlocal means is put forward. Compared with local filters, the nonlocal filters have good performance in image denoising. We investigate a nonlocal means algorithm to denoise a difference image. The similarity measurement between two patches is a crucial problem in nonlocal means. In order to adapt to synthetic aperture radar images, the ratio similarity measurement is applied in nonlocal means.