Remote Sensing Image Registration Algorithm Based On Improved SIFT And Improved K-means

With the wave of the third technology revolution,Modern science and technology have been developing rapidly,and the rapid development of aerospace technology is also promoted.Human could obtain increasingly rich information on remote sensing images.The development of science and technology has also enabled the continuous upgrading of hardware devices,and the continuous enhancement of sensor resolution is a typical example.The progress of science and technology has led to the development of human research.Increasing number of researchers focus on the high-resolution images with a spatial resolution of approximately 1 m.Remote sensing image registration is a key step in remote sensing image processing.When images of the same scene were acquired by different sensors,from different viewpoints or at different times,there will be serious geometric deformation of remote sensing images,which will bring difficulties to image processing.Remote sensing image registration is to restore remote sensing images with geometric distortion before processing all kinds of remote sensing data.The specific process is to use spatial geometric transformation to transform two or more different types of remote sensing,so that the corresponding regions representing the same location in the image are mapped to the same coordinates,then matched or superimposed.Automatic registration of remote sensing images has become a hot topic in the field of current research.Aiming at the problems such as false matching points and large volume remote sensing image registration by SIFT algorithm,a remote sensing image registration algorithm based on improved SIFT and improved K-means is proposed.The specific process of the algorithm is as follow: first,feature points were extracted using the Harris corner detection algorithm to obtain the matching points to execute the coarse registration.The low scale space feature points were extracted after down sampling the images and extracting the interesting region of two images.The improved K-means clustering algorithm was used to classify the feature points.Every class is a region.Furthermore,feature points in each region were used to obtain the Delaunay triangulation,and then calculated the similarity between regions of both images to select pairs of triangles which the similarity greater than threshold.Finally,the fine registration was achieved by precise feature points.The experimental results indicate that the proposed method is effective.