Research On Parallel Classification System Based On The Remote Sensing Image Data Of The Latitude And Longitude Grid

Remote sensing im age classification is an important branch of image tar get recognition,widely used in agriculture and forestry management,land planning,marine observation,environmental protection,disaster m onitoring and estim ation of national economic construction field.In recent years,with the rapid developm ent of domestic and foreign remote sensing platform and sensor technology,remote sensing data daily gets has reached 10 TB.The rapid increase in the a mount of data,innovate organization and storage of the remote sensing image data.The current popular processing method is to form image "tiles" by using im age normalized segmentation,and realize cloud storage of tile im age by using distributed storage technology.Segmentation of images further intensifies the growth of data size,b ut the curr ent remote sensing im age classification systems are more suitable for the single view rem ote sensing image data of independent file or ganization and centr alized storage m anagement,processing performance is low,automation degree is low,can not meet the massive tile image data fast,real-time,efficient information excavation.In order to explore the new business processing model of tile remote sensing image classification in large data environment,and improve the business level of processing,the following work has been m ade in this paper:(1)The "five tier fifteen level" image segmentation organization standard based on the latitude and longitude grid is studied,and im age data of "Five-layer Fifteen-level" tile are us ed as the processi ng object of parallel classification system,then the organization,storage and management of tile images in the cloud database are discussed.(2)The file or ganization of supervised classification samples based on tile im age data is explored,raster data is used as the sample data structure,the metadata table and related information table f or the sample ar e designed,distributed sam ple database is implemented based on tile cloud database technology.(3)Combined with the current popular supervised classification algorithm,a distributed parallel processing model is proposed,and the multi task driving mechanism of this model is designed and im plemented.The design of the system architecture and function modules of parallel classification syst em is finished,and with the help of.Net and DotSpatial platform complete the deve lopment of each m odule,the parallel classification system is realized.(4)The hardware and software environm ent of high perform ance cluster is built,experimental data preparat ion was carried out,the ef fects of cluster nodes and classification algorithms on the perform ance of parallel classification system s are discussed,the classification accuracy of each algorithm in the system is also evaluated.Through the above key technical problem s,this paper im plements a high performance parallel classification system based on the im age data of the latitude a nd longitude grid.After experimental verification and running test,the results show that :The system can greatly improve the processing speed of the batch tile image data in the premise of ensuring the accuracy of classification,has obvious advantages compared with the traditional processing methods.