SAR Image Change Detection Based On Super-Pixel And Saliency Detection

Remote sensing image change detection technology is analyzes the change of the surface according to the difference of two or more images,which were collected in different time periods in the same ground scene.The technology is a research hotspot in the field of remote sensing images.Synthetic Aperture Radar,sort for SAR,is one kind of remote sensing images.SAR images are formed by active microware reflection,escaping from limits of light and time conditions,which makes SAR image more suitable for change detection.Therefore,it's very important significance and broad prospects for studying the SAR image change detection.Traditional SAR image change detection is based on pixel,which don't take images' spatial information into account.Therefore,the paper presents a SAR image change detection based on super-pixel segmentation to achieve this goal.Inspired of the saliency detection about optical images,the paper proposes a saliency detection method for SAR image change detection.The method not only accurately detects the range of changes,but also reduces the effect of speckle noise.The surface area varies in size and presents a multi-sacalle distribution.Therefore,the image scale is also introduced into the image change detection.In order to ask the question that how to effectively enhance the precision of SAR images' change detection,the paper proposes two methods which are used by SAR images change detection:(1)A method for detecting changes of SAR images based on super-pixel and saliency detection is put forward.Traditional SAR image change detection is based on pixel-based features,which don't take images' spatial information into account.It has a great impact on the accuracy of the detection results.So the paper uses super-pixel features instead of pixel features,which can obtain good edge information and make use of images' spatial information.Super-pixel can effectively reduce noise and imporive the speed of operation.In order to quickly and accurately obtain the difference map,a saliency detection method is presented,which can roughly detect the scope of change.It improves the accuracy of the change detection results.Experimental results demonstrate that this approach effectively enhances the image change detection accuracy and remains the better edge information and details.(2)A approach for detecting changes of SAR images based on multi-scale and mutual saliency detection is presented.Unlike traditional saliency detection methods,the paper uses two images to gain the saliency map.We use the super-pixel segmentation technique to segment the two images separately.Two roughly changed saliency map are obtained by using the mutual saliency detection method.Finally,we merge the two saliency maps to get a new saliency map.The purpose is to make betteer use of the change information of the two images.Further,we use multi-scale analysis to improve the accuracy of the change detection.Experimental results show that the method can effectively improve the precise of change detection results and preserve the image edge information and details and so on.