Research On Seamless Stitching Technology Of Geostationary Ocean Color Radiometer

China is a big ocean country.Changes in the ocean are closely related to China's national defense,economy,and people's livelihood.China has used polar-orbiting ocean satellites to observe oceanographic information for more than ten years.Since polar orbiting satellites pass through the same place on the earth at the same time in a day,the revisiting time of polar orbiting ocean satellites are used to observe specific areas is more than one day.However,the tides,currents,pollution,and military activities of the ocean often change greatly during the day.Therefore,improving the time resolution of ocean satellites is extremely important.In the project,geostationary ocean radiometer is used to increase the time resolution of ocean satellites.Limited by detector size and the field of view of the optical system,the ra-diometer uses a two-dimensional pointing mirror combined with an array detector to expand the observation range by stitching the sub-images.When the two-dimensional pointing mirror rotates around two axes,scanning nonlinearity and image rotation error are introduced,result-ing in the omission of geographic information,while the place could be seamlessly coverd if there's no errors.My thesis focuses on how to seamlessly stitch subimages of geostationary ocean radiameter.The main contents and innovations of the paper are as follows:1.The imaging principles,scanning trace and sub-image features of the two-dimensional pointing mirror were analyzed.The relationship between the scanning angle change and image cover condition was given.The conclusion that a positive pitch angle was better to observe the northern hemisphere was proposed.Two methods for judging scan miss-ing are proposed:the grid method and the geometrical method.The two methods are compared and the results of scan missing judgement with different pointing angles and different interval angles are given.Analyzed the relationship between scanning interval and coverage ratio,the influence of earth curvature on the imaging,and proposed a seamless scanning design process.2.According to the corresponding relationship between object and image of the pointing mirror and the interpolation algorithm,an algorithm for image rotation elimination is proposed,and verified by using feature points slope difference between adjacent sub-images.Finally,the image rotation elimination results and their verification results are given.It is verified that the image rotation elimination algorithm proposed can reduce 39%of feature points slope difference between adjacent sub-images.3.In order to obtain registration parameters of remote sensing image stitching,a global and local-based registration method can be used.For the model machine images obtained in this paper,a local-based registration method is used.The local feature extraction methods of remote sensing images mainly include straight line feature extraction and point feature extraction.The Hough algorithm and LSD algorithm perform well in the straight line feature extraction.However,since the feature extraction of the straight line requires obvious line features which the remote sensing images always lack of,the feature extraction algorithm is selected in SIFT algorithm,SURF algorithm,and ORB algorithm.The principle of three feature point extraction algorithms is analyzed,and the method of image registration is given.Use the image of the stationary ocean radiometer images to perform feature point registration.The registration results of adjacent images are given,and three different feature point ex-traction algorithms are compared by the registration results.Finally,the ORB operator,with good registration performance,short computing time,and small memory using,is selected for feature extraction4.On the basis of image registration,erroneous matching is eliminated by RANSAC algorithm,and the homography matrix of adjacent image is obtained.After image transformation,the stitching gap is eliminated by the distance-weighting method,and the result of sub-image mosaic of the radiometer image is given.Combining the image rotation elimination algorithm,the comparison of mosaic results before and after image rotation elimination is given,and it proves the position accuracy of the two-dimensional pointing imaging and remote sensing instruments is improved.This subject can obtain the maximum pitch and azimuth angle intervals covering the hemi-sphere,and can seamlessly stitch through the judging method of scan missing,giving theoretical basis for setting the motor angle,thus ensure the time resolution of the radiameter.In the image stitching,the ORB operator with a stitching time of about 2s was raised,which provided the possibility of real-time image stitching on the satellite...