Research On UAV Marine Remote Sensing Image Processing And Stitching Algorithm

With the development of UAV technology,remote sensing technology based on UAV platforms has become the frontier of marine remote sensing.Compared with satellite remote sensing,UAV near-ground remote sensing has the advantages of high image resolution,flexibility,economy and less influence by the operation cycle,etc.However,the field of view of UAV is small,and the stitching process of UAV remote sensing images is needed,and the pre-processing of remote sensing images will ensure the quality of subsequent image stitching.Therefore,this paper researches on the pre-processing and stitching algorithm of UAV remote sensing images of ocean field of view,and the main work accomplished in the thesis is as follows.Firstly,a morphology-based multi-structured morphological filter is designed for the problem that the marine remote sensing images have little detail information and the commonly used filtering methods often lead to excessive smoothing of the images and loss of the edge detail parts of the images.The method introduces multi-scale multi-structure elements for image filtering,which increases the image quality while preserving the original structural elements of the image and retaining more image details for later feature extraction.Experimental results show that compared with the median filtering,which is also a non-linear filtering method,the improved algorithm in this paper improves the edge intensity,average gradient,information entropy and image sharpness indexes by 7.735%,20.36%,1.093% and 30.45% respectively.Secondly,to address the problem that the commonly used random sampling consensus precise matching algorithm cannot guarantee the dispersion of matched feature points and the difficulty of the algorithm to judge when the number of interior points are detected to be the same,the GA-RANSAC algorithm was obtained by introducing the gene algorithm to improve the algorithm,which adds the dispersion of interior points to the fitness function of the gene algorithm to make the selected feature matching points more discrete.Finally,in order to further improve the accuracy of UAV remote sensing image stitching,the global transformation model between images is optimised in combination with the existing elastic local alignment algorithm,thus further improving the performance of the algorithm in this paper.The effectiveness of the improved algorithm in this paper is verified through ablation experiments and comparison experiments and evaluation with other existing algorithms.The experimental results show that the improved algorithm in this paper can better align the feature points of the image to be stitched,effectively improve the image overlap area ghosting problem,and the information entropy of the stitched image is improved by 0.62%,5.06% and 9.93% respectively compared with the RANSAC global projection transform,APAP and AANAP algorithms.In summary,this paper thoroughly investigates the image pre-processing and stitching algorithms of marine remote sensing images,provides a set of stitching processes and algorithms applicable to UAV marine remote sensing images,and provides a technical basis for the implementation of UAV marine remote sensing monitoring system.