Research On Small Celestial Body Feature Point Matching Method For Landing Segment

The small celestial body image taken by the optical camera has the characteristics with many details,high resolution.It also has more detailed image features than the fly-around segment image.So it is widely used for optical autonomous navigation.The small celestial body image is affected by the descent speed of the detector and the imaging time of the optical camera,which make the image blurred,and have a certain influence on the validity of the image processing method.In the process of the small body image processing,the image feature points matching is the core of the image processing method.The small body image feature points matching accuracy is affected by the image quality and feature points matching criteria.The research on the high quality of image restoration method and the high precision of the image feature points matching method,is very important for improving the performance of the small body detector image navigation in the landing segment.In the research of the small body image processing method,the PCA-SIFT algorithm is a classical algorithm of the image feature points extraction and matching,which can obtain the high accuracy of the image feature points in theory.But the Euclidean distance is used as the similarity matching criteria,which makes the different dimension index equal to the same scale level.The result is that the matching precision reduces.In this paper,according to the study of the small body image feature points matching method based on the traditional PCA-SIFT algorithm and the image feature of the landing segment,we improve the algorithm from optimizing the small body image quality and increasing the image feature points matching accuracy.(1)In order to improve recovery quality of the blurred image,a method based on spectrum features is proposed.By this algorithm,the small celestial body image noise is first eliminated by the NSCT filter.Then,the simulated annealing algorithm is used to optimize the image spectrum features to obtain the point spread function.The fuzzy degeneration model is established.Finally,the blurred image is recovered by the Lucy-Richardson algorithm.The recovery quality is evaluated by the average gradient and the Laplace operator.(2)In order to improve the matching precision of the small body image feature points,we propose a method which combines PCA-SIFT with correlation coefficient.Firstly,the PCA-SIFT is utilized to extract restored image interest points.And then,the correlation coefficient is used as similarity measurement,which can filter image interest points.Finally,the RANSAC algorithm is used to eliminate wrong and repeated matching pairs.In this paper,the small body image of landing segment is obtained from the image database provided by the National Aeronautics and Space Administration.We simulate the above two methods.The simulation results demonstrate that the image recovery quality by the proposed method improves greatly compared with the traditional method.It lays the foundation for image feature point matching.Compared with the traditional PCA-SIFT algorithm,the correlation coefficient can reduce wrong feature points,and improves image navigation performance of the detector.