The Study Of SAR Image Change Detection Method Based On Target Detection

Image change detection techniques aim to detect the changes in the same area over time. They are very important in remote sensing techniques and are widely used in many fields such as environment, agriculture, water conservancy and military. Synthetic Aperture Radar (SAR) is good information resource for change detection because of its all-time/all-weather capability. Change detection with SAR image has wide application prospect.An accurate registration is a very important precondition for most of the present change detection methods, but the technology of image registration is still a hard work in the area of image processing especially for SAR image, so it is rather difficult to find a good SAR image registration method, which should have both strong adaptability and high accuracy. This will restrict the application of change detection for SAR image to some extent.Considering the problem of registration accuracy for the present change detection methods, and according to the characteristics of man-made target in SAR image, a new change detection method with SAR image is presented in this paper, which i s b ased o n t arget d etection. In t his m ethod, t he c hange d etection p rocess i s divided into two steps, that is target detection and change detection. First, the man-made targets in the images will be detected; second, we will compare the similarity between the relative position relationships of the man-made targets which have been detected from the two images, then the change of images will be decided from this result. As the change detection process which is based on target detection is mainly dependent on the relative position relationships of the detected targets, the requirement to the accurate registration can be greatly reduced.In the target detection step, a kind of improved two-parameter CFAR detector is presented. It has a better performance than the common CFAR detector in the complex environment such as multiple targets environment. In the change detection step, a kind of mathematical morphology closed operation is used. As it has a better ability of smoothing target edge than the general mathematical morphology closed operation, it is fitter for extracting the character points of targets.For all the researches in this paper, the SAR images of Beijing in different times are used to validate. The experimental result shows that this method is capable of detecting image changes with rough registration accuracy.