Optical And SAR Water Image Registration And Change Detection

In sudden natural disasters,in order to understand and relieve the disaster immediately,Synthetic Aperture Radar(SAR)system has become the preferred equipment for remote sensing to obtain the damaged area,because of its all-day and all-weather characteristics.The registration and change detection of SAR image and optical reference image have become the research hotspot in remote sensing image processing.However,in the optical and SAR image registration,the common features are difficult to obtain because of different radiometric and geometric properties.Meanwhile,the large scene images are too time-consuming to process in real time.In order to avoid the difficulty of registration and change detection caused by anisotropy of anisotropic images,large and uniform waters are selected as their common features as the basis of research.In this paper,we mainly study that how to realize real-time matching of SAR water images with pre-stored optical reference maps and establish stable feature registration methods,and focus on realizing optical and SAR water change detection.The main work and contribution are as follows:Firstly,in order to improve the accuracy and efficiency of the rough matching between optical and SAR water images,a fast matching method is proposed based on the connected region of common waters.In view of the traditional matching method,it is difficult to match the matching between optical and SAR water images based on the similarity of the image grayscale.Thus,a similarity measure method based on the minimum convex polygons and the minimum bounding rectangle is proposed to realize the coarse matching between optical and SAR water images.It solves the difference of the heterogenous image and avoids the pixel traversal of the traditional matching method.Finally,it realizes the fast matching.Secondly,in order to improve the accuracy and efficiency of accurate registration of optical and SAR water images,a fast optical and SAR water image fine registration method is proposed based on Second Otsu image segmentation and scale-invariant feature transform(SIFT).The Second Otsu segmentation is used to achieve more accurate water segmentation.It takes the waters as a distinctive feature and removes the difference caused by the the strong scattering in the non-water area of SAR images and the multi-band reflection of the optical image.Then,the improved spatial consistency criterion is used to eliminate the mismatch features and by and improve the accuracy of the matching points so as to prevent mismatch of the RANSAC algorithm caused by too many false matching points.The experimental results show that the proposed method can effectively increase the accuracy of optical and SAR image registration and the computation efficiency.Thirdly,a water object classification extraction method based on SLIC superpixel segmentation and combination is designed.Superpixel region merging is used to preserve typical features of ground objects and avoid water holes and discontinuities caused by speckle noise in SAR images.Then the support vector machine is used to extract the water object.Then,aiming at the problems of geometric error,registration error and segmentation classification error caused by the difference of optical and SAR images,a water area change detection method based on FCM clustering and multi-scale decision fusion is proposed.The logarithmic ratio method and FCM clustering are applied to the acquisition and analysis of differential maps.Finally,the final change information is obtained through multi-scale decision fusion.The experiment shows that the algorithm can be used to detect the change of optical /SAR water well.