The Research On Remote Sensing Image Registration Algorithm Based On RANSAC And Entropy Constraints

In recent years,the feature extraction and matching of remote sensing images is a hot research topic.Therefore,how to improve the matching accuracy of remote sensing image under the premise of ensuring the efficiency of the algorithm is the current research difficulty.In the registration of Remote Sensing Image,the main steps are feature extraction,feature description,and feature matching.In this paper,representative research is performed on the feature extraction algorithm and matching algorithm in remote sensing image registration,and the corresponding improved algorithms are proposed,respectively.A commonly used method in the matching stage is the random sampling consistency RANSAC algorithm,which removes the mismatching points caused by the matching phase by estimating the model parameters.For the traditional RANSAC algorithm,because there is no upper limit in the number of iterations.A large number of samples and solve the homography matrix will bring a lot of disadvantages,such as increasing the number of matching operations,increasing the time-consuming matching,and the accuracy of the single response matrix obtained by this algorithm is not high.For the shortcomings and problems of the traditional RANSAC algorithm,this paper based on the SURF feature extraction algorithm,the algorithm has the following characteristics,First,reducing the number of iterations to improve the efficiency of the algorithm,Then,quickly abandoning unreasonable matrices to reduce the time of homography matrix verification.Finally,updating iterations to improve the accuracy of homography matrices.The experimental results show that the algorithm proposed in this paper not only improves the accuracy of the homography matrix but also reduces the execution time of the algorithm.Aiming at the problems that KAZE algorithm extracts feature points of low accuracy and mismatching for remote sensing this paper proposes a preprocessing algorithm to accelerate KAZE feature extraction.The new algorithm first proposed to preprocess the remote image based on entropy constrained and KAZE feature fxtraction.The method first uses a non-overlapping sliding window to traverse theremote sensing image and segments window area,and the entropy of the segmented window area is sequentially calculated.Then,according to the histogram formed by the obtained entropy,an appropriate threshold is selected to retain the local area of image with high entropy for feature extraction of the KAZE algorithm.Finally,the RANSAC algorithm is used to remove mismatches to optimize matching results.Experiments on the SPOT,GH-2 satellite data indicate that compared with the KAZE algorithm without coupling proposed algorithm,the accuracy of the KAZE algorithm coupled with the proposed algorithm is improved by 0.2%,0.3%,and the performance time of algorithm has been decreased by 70%,53% respectively.