Research On Hight-resolution Remote Sensing Image Mosaic Technology

The rapid progress of remote sensing technology has promoted the development of human cognition,and has been applied in many fields such as land and resource protection,urban regional planning,disaster prevention and relief.The acquisition of high-resolution remote sensing images enables people to view the world more precisely and accurately.However,the increase in the resolution of remote sensing images has severely limited the coverage area of single-scene images.At the same time,due to weather conditions images,it is difficult to obtain large-scale high-quality images at the same time in some cloudy and rainy areas,and it is necessary to conduct mosaic research on images with different time phases.Aiming at the characteristics of high-resolution remote sensing images,in order to meet the needs of large-scale research,this paper combines image processing algorithms to study the automatic mosaic technology of high-resolution remote sensing images.The main research work and innovations include:(1)Understand the development status of remote sensing image mosaic technology at home and abroad,and determine the key research directions: Mainly research on remote sensing image registration and mosaic line extraction.Introduced the current common methods of remote sensing image data processing,such as image denoising,geometric correction,panchromatic multi-and multi-spectral fusion,uniform light uniform color and other preprocessing.(2)Compare the advantages and disadvantages of Harris corner detection,SIFT and SURF feature extraction algorithms in remote sensing image registration.In order to ensure the accuracy of high-resolution image registration and improve the efficiency of registration,improvements are made on the basis of the SIFT algorithm.Use the neighborhood tracking method to sample the effective area contour of the orthoimage and compress it,and perform geometric calculation on the effective area polygon vector to obtain the overlapping image area.SIFT is used for registration in the overlapping area of the image,which reduces the range of feature point extraction and matching,reduces the registration time,reduces the occurrence of false matches,and improves the quality of registration.(3)A mosaic line extraction technology based on multi-scale segmentation is proposed.Cluster the obvious feature areas to generate dense superpixels,and obtain the texture information of the features in the image;Then,the adjacent regions are merged by continuously increasing the regional dissimilarity threshold,and the scale set model is used to record the region merging process;At the same time,the optimal segmentation scale is determined according to the local variance of the spectral characteristics and the Moran index to solve the problem of over-segmentation;Finally,the A* algorithm is used to find the best mosaic line on the segmentation path.Experimental results prove that this method effectively solves the problem of mosaic lines passing through obvious areas such as buildings,farmlands,rivers,etc.,and reduces splicing traces.It can be widely used in highresolution remote sensing image mosaic and mosaic,and has practical significance for remote sensing image automatic mosaic.