Research On Remote Sensing Image Mosaic Method Based On Improved Matching Strategy

Remote sensing image mosaic technology has very important research significance.The stitching technique is the main method used to obtain large-scale remotely sensed images.This is not only conducive to the far-reaching development of the aerospace imaging technology in China,but also only plays an indispensable role in various fields such as agriculture,forestry,water conservancy,and military defense.In stitching technology,image registration and fusion are the key steps that occupy the main position.Among them,image registration directly affects the quality of stitched images,which is the most important step.Therefore,scholars have been focusing on image registration algorithms with high accuracy and fast speed.In addition,image fusion also undertakes important work.The fusion algorithm can reduce the appearance of stitching seams,smooth the possible color and brightness differences,and effectively improve the visual quality of the image.The main research content of this dissertation is the exploration and improvement of low error matching strategy and image mosaic fusion algorithm.And finally achieve a better visual mosaic effect of remote sensing images.The main work of this article is as follows:(1)Aiming at the common noise in remote sensing images,this dissertation uses remote sensing image guided filtering combined with weighted median filtering.First,the Gaussian weighting function is used to optimize the median filtering to improve the disadvantage of the median filtering to treat neighboring pixels equally when processing noise.Then combined with the ability of guided filtering to preserve details.Finally,a filtering method with better denoising ability is formed.The experimental comparison with common filtering algorithms proves that the filtering algorithm used in this dissertation has a good denoising effect.This plays an important role in the further research of remote sensing image mosaic technology.(2)Aiming at the situation that the stitching positioning of remote sensing image mosaic is inaccurate and the image stitching seam is obvious,this dissertation proposes a remote sensing image stitching algorithm that combines the second-order gradient and golden section cyclic correction.First,a KNN matching strategy that incorporates the second-order gradient is discussed,and the second-order gradient description that can reflect image texture information is incorporated into the matching algorithm.This can reduce the mismatch rate and achieve accurate calculation of the image relation conversion matrix.Then based on the best suture search principle,the golden section method is used to cyclically optimize and correct the pixels in the specified area to achieve the precise positioning of the best suture.Finally,the image mosaic is completed by combining the fade-in and fade-out fusion algorithm.Experiments show that the proposed algorithm can improve the correct matching rate and achieve seamless fusion and stitching of remote sensing images.(3)Aiming at the high matching error in remote sensing image stitching and the sudden change of brightness in the overlapping areas,this dissertation proposes a nonlinear S-type remote sensing image stitching algorithm under a two-way matching strategy.First,this dissertation uses a two-way matching strategy to reduce the traditional one-way matching error.The RANSAC algorithm is used to reduce the mismatch rate again to construct a lowerror two-way matching strategy.Secondly,based on the law of human eye visual perception curve and plant growth curve,explore an S-type weighted fusion factor that can adapt to human eye perception to form a smoother fusion algorithm.This solves the problems that direct average fusion and fade-in and fade-out fusion cannot well eliminate stitching seams and improve the brightness difference of overlapping areas.Through subjective and objective experimental analysis,it is proved that the stitching algorithm proposed in this dissertation can reduce the mismatch rate,has a better fusion and stitching ability,and can achieve the fusion and stitching of remote sensing images without obvious brightness changes.