Research On Key Technologies Of High-performance Remote Sensing Image Tile Service

In recent years,with the development of drone aerial photography and satellite remote sensing image technology,major research institutions and Internet companies have been able to obtain a large amount of remote sensing image data every day.At present,remote sensing images have become basic data widely used in scientific research,Internet applications,national defense and military fields.At the same time,to better realize the sharing and utilization of remote sensing image data,online and servicing are the development trends of remote sensing image applications,and map tiles have gradually become the main form of distribution of remote sensing image data on the Internet.However,with the advancement of modern information technology,the amount of remote sensing image data has increased exponentially,and traditional remote sensing image online service methods have gradually exposed problems such as the inability to efficiently process large-scale image data and untimely service response.How to improve and enhance the performance of fast remote sensing image services and provide better remote sensing image services for the majority of Internet users is a hot research topic in the field of geographic information today.This article focuses on the hotspots and difficulties in the existing remote sensing big data online tile service technology,namely,the rapid tiling method of remote sensing images,the storage management method of raster tiles for high concurrent access,and the tile for user operation habits.Research on film service methods.The main research contents are as follows:(1)A fast grid tile generation algorithm that takes into account task dependence is proposed.Among the traditional grid tile pyramid generation algorithms,there is a parallelized tile generation algorithm based on synthesis from high-level tiles to low-level tiles.It is widely used because it can skip the step of generating image pyramids and directly generate tile pyramids.However,the problem of task dependency between upper and lower tiles in this algorithm causes processes to wait for each other at each level,which causes a lot of waste of time.This paper designs a multi-level task division algorithm,which reduces the impact of task dependence as much as possible,reduces the time cost of tile generation and the cost of process creation between each level.It further improves the speed of grid tile generation.Experiments show that this algorithm improves the pyramid generation speed by about 30% compared with the original algorithm.(2)A raster tile organizational management for mass storage and high concurrent access is proposed.Given the poor performance of tile management caused by a large number of small files in the traditional raster tile storage method,this paper proposes the use of map tile file aggregation storage technology,which reduces the disk storage and read-write overhead of map raster tile metadata.Based on the research on the organization and management framework of map tiles for high concurrent access,this method considers the spatial proximity of map tiles on storage,and proposes a file merging strategy with spatial characteristics of map tiles.This strategy further improves the overall read performance of raster tile data.In addition,the organization method proposed in this paper also takes into account the spatial proximity of map tiles in storage and combines the file merging method with the spatial characteristics of map tiles to store spatially adjacent tiles in a data set file.It further improves the overall read performance of raster tile data.Experiments show that this method saves about 28% of storage space compared to traditional file systems,and the read speed is 5-20 times faster than traditional file systems and relational database organization methods in different situations.(3)A grid tile service method based on real-time tile prefetching is proposed.Traditional raster tile prefetching strategies mostly analyze the temporal and spatial characteristics and access characteristics of tile data and use big data to determine the cache weight and prefetch order of tiles,often ignoring the predictability brought by the user’s user habits.Based on the user’s habits,this paper improves the traditional prefetching strategy by combining the depth-first strategy and the breadth-first strategy and designs a tile prefetch strategy for individual users to operate instantly,which makes full use of the user’s access to tile services.The waiting time is prefetched in advance for the tiles that the user may access in the next step,and the number of disk operations is reduced through the merge request strategy,which relieves the server pressure in the case of high concurrent access and further improves the user’s online remote sensing image service experience.