Research On Rapid Tile Pyramid Construction Technology Of Large-scale Image Data Set

With the rapid development of computer technology and remote sensing technology,the resolution of remote sensing images is getting higher and higher.At the same time,the scale of data is also greatly increased,and a single high resolution remote sensing image has reached hundreds of GB of data.The in-depth development of WebGIS makes the global real-time sharing of geographic data become an important goal of the development of geographic information industry.The relative contradiction between the increase in the size of geographic information data and the rapid response of geographic information services has led to the requirement of rapid map data tiling.At present,the improvement of hardware performance and the development of various types of highperformance computing provide a new solution for realizing the rapid tiling production of large-scale image data sets.How to use the parallel computing and distributed computing to realize large-scale tile pyramid construction become a focus of GIS-related researches.Based on the practical application,the paper looks for an efficient and stable largescale image data set tile pyramid rapid construction technology,achieving a reliable image data set tiles pyramid rapid construction through the MPI multi-process parallel way.Specifically,the following three aspects were solved.First,multi-source image data projection transformation and coordinate transformation.Large-scale image data sets often includes a variety of image data with different geographic coordinate system and projection coordinate system.To build tile pyramid with these image data,firstly it is needed to convert them to the required coordinate system and then other operations are followed.The implementation of the Bursa model and the control point based matrix conversion method makes it a feasible way.The process that fist converting the four vertexes and then resampling filling optimizes the conversion efficiency.The second is to achieve image data mosaic and slice parallelization.The process of allocating the calculation process to multiple processes is carried out by the task division method of the wheel method.The parallelization method greatly improves the efficiency of the algorithm.The equalization of the tasks makes the overall parallel efficiency tend to be optimal.The third is to find a reasonable way to manage and store the tile pyramid data.There are a large number of tile pyramid files and it is obviously needed to find the right way to store,manage,and apply a wide range of tile files.By comparing the advantages and disadvantages of different storage models,we can find the best storage model under different application scenarios to realize the efficient management and application of tile data.Through the experimental verification,the rapid tile pyramid construction technology of the large-scale image data set proposed in this paper can improve the efficiency of the algorithm of other existing software or tools under the premise of correctness of the algorithm.Besides,the performance is stable and the algorithm has strong portability and expansibility.The parallel tile pyramid construction technology can be applied to various kinds of application with requirement of rapid data tiling of remotesensing image.