Research On The Generation Of Grid Seamline Network Considering Cloud Coverage

In the research and application of remote sensing technology in large areas,image mosaicing technology is indispensable.The seamline network-based mosaicing technology is independence of image processing order,so it has been studied and applied widely.However,most of the existing seamline network generation methods are based on vector calculation,which result in the consequence that they can't address complex overlapping areas and they are unstable.In addition,for remote sensing images,the cloud is ubiquitous,which will cause the loss of information under the cloud and should be avoided as much as possible.Therefore,this paper studies the grid-based seamline network generation method that can adapt to complex overlapping areas and large data volume firstly.In addition,this paper studies the seamline network optimization method considering cloud coverage.The grid-based seamline network generation method and the seamline network optimization method considering the cloud coverage are combined by the effective mosaic polygons to generate the minimum cloud coverage mosaic image.The main research work of this paper is as follows:(1)This paper proposes a grid-based seamline network generation method.From the perspective of the grid,the method simplifies the complex overlapping problem between multiple images into many overlapping problem between two intersecting images,and uses the improved region growing algorithm to process the overlapping regions of arbitrary shapes between the two intersecting images.The improvement points mainly focus on that the improved region growing algorithm uses the boundary of the overlapping area as growing points.According to the intersection points of the effective area boundary of two intersecting images,the boundary of the overlapping area can be divided into two parts,which are used as the growing points of the region growth,and the two parts carry out the regional growth simultaneously.In addition,the region growth is only carried out in the overlapping area.Finally,the junction of two growing areas is the seamline.Experiments show that the grid-based seamline network generation method can process complex overlapping areas and large data volume,and generate mosaicing image without redundancy and seamlessness.(2)This paper proposes a quad-tree iteration-based seamline network optimization method.The grid-based seamline network generation method only considers the positional relationship between images,and generates the initial seamline network,but the seamline passes through the cloud area,and the mosaic image is covered with clouds are difficult to avoid.Therefore,this paper proposes a seamline network optimization method considering cloud coverage.This method fully considers the spatial distribution of the cloud coverage areas,and uses the quad-tree iterative method to optimize each cloud area by refining the effective mosaic polygons,which ensures that each cloud area is optimized with as few images as possible,and finally generates a minimum cloud coverage mosaic with better radiation continuity.Experiments show that this method can ensure the radiation continuity of the image and get higher efficiency compared with the pixel-based mosaic method considering cloud.