The Study Of Change Detection Technology In Multi-temporal SAR Images

Change detection technology acquire change information of interesting objects, such as the shape, situation, quantity or other aspects, mainly through detecting the differences of intensity or local texture information between multi-temporal images. Synthetic Aperture Radar(SAR) images are affected by speckle noise seriously because of the unavoidable reasons in radar imaging. It is difficult to improve the performance of change detection technology. In this paper, we mainly study the change detection technology with SAR images and the performance evaluation of it.Firstly, the development and the situation of the change detection technology and some problems existing in this field are introduced. With these problems, the pre-process in change detection with SAR images is introduced in chapter 2, which includes image registration, image speckle filtering and image intensity match. In the end of this chapter, a new method in reducing effect of image misregistration in change detection with SAR images is presented. In chapter 3, the change detection method based on image intensity differences are mainly introduced. With the difference image, some classic image threshold auto–select methods are recommended, such as circulation segmentation, maximum deviation between different classes, threshold auto-select method based on the best entropy, threshold auto-select method based on KI algorithm and threshold auto-select method based on EM iterative algorithm. In chapter 4, a novel performance evolution method of change detection based on simulative SAR images is presented. Through segmenting SAR images into background and object chips and the combination of them, simulative SAR images which have very similar characteristics with the real SAR images are acquired. With the simulative images, it is easy to get the true ground situation change image. Different change detection methods based on difference image are used to detect the simulative images. Comparing the detect results with the true ground situation image, the performance of all kinds of methods can not only be evaluated, but also make the evaluation results more believable.Finally, in chapter 5, the main work of this paper is summarized and the shortages of our work are presented. With indicating the prospect of change detection technology in SAR images, the direction of our further work are pointed out.