Research On Mosaic Algorithm Of Agile Satellite Remote Sensing Image

Remote sensing image mosaic is a kind of image processing technology which combines a set of remote sensing images from different sensors,different time and different perspectives into seamless images with high resolution and wide field according to their overlapped parts.The technology mainly resolves the contradiction between resolution and field of view and is an indispensable part of remote sensing image processing.It has been widely used in remote sensing image analysis tasks such as change detection,target recognition and map updating.As an important direction of satellite technology development in recent years,agile satellites have the advantages of fast maneuvering speed and high degree of freedom.However,such features also put forward higher requirements on accuracy,speed,memory and robustness of mosaic algorithm.Therefore,this paper improved the classic algorithm of image mosaic based on the mosaic constraints of agile satellites and proposed a new mosaic algorithm.We carries on the study from three aspects: image preprocessing,feature matching,image transformation,and fusion.The main work can be summarized as following:(1)This paper presents an improved preprocessing algorithm based on histogram equalization for the problem of noise interference and gray distribution.The algorithm can effectively reduce the influence of noise,improve the gray distribution of the image,and reduce the disregistrated rate caused by forest area and calculation of algorithm,and make preparation for subsequent feature matching.(2)Under the restrictions of agile satellite,SIFT and SURF algorithms are selected for their strong robustness,and are compared and analyzed in the matching performance of remote sensing images.Given the better performance of SURF algorithm,this paper presents an improved matching algorithm based on it.The presented algorithm can fully meet the requirements of agile satellite in the accuracy,computational speed,and robustness.(3)The paper chooses RANSAC algorithm to optimize the matching points,calculate the image transformation model,and achieve the image projection transformation and initial image stitching.Three fusion algorithms,gray-weighted smoothing,median filter,and wavelet transform,are used to remove the seam,and the effects of fusion are compared and analyzed.Seamless mosaic of agile satellite remote sensing image is finally realized.