Research And Implementation Of Remote Sensing Image Processing Cloud Model And Fast Terrain Extraction

With the development of the technology of sensors,Satellite imagery acquisition technology has made considerable progress.The incensement of bands and resolution of satellite imagery causes exponential growth of image data,resulting in a large amount of remote sensing data.How to effectively enhance the ability of remote sensing data processing has become an urgent problem.The tremendous development of high performance computing,especially cloud computing technology,provides a new way to overcome that problem.However,because the existing cloud computing platform is not designed to process the remote sensing imagery,deficiencies inevitably exist in the process.Starting from exploring the nature of the cloud computing technology,based on the characteristics of general remote sensing data processing and inspired by the basic idea of divide and rule in high performance parallel computing,a Cloud Computing Remote Sensing Image Processing model(C-RSIP)is designed in this paper.A Extend Twice Match(ETM)method for fast terrain extraction and an adaptive noise cancellation algorithm based on density analysis and a hierarchical weighted grid fitting algorithm are presented for professional remote sensing data processing;For application,a ZY-3-APGS system based on C-RSIP has been developed and equipped for the Satellite Ground Data Handling System(GDHS)in China Centre for Resources Satellite Data and Application(CRESDA)and a TH-1B-APGS system based on C-RSIP has been developed and equipped in the TH Satellite Ground Data Handling Centre.The main contents are as follows:(1)A cloud computing model C-RSIP is designed for remote sensing data processing.The model C-RSEP is strictly in accordance with three service architecture design:A multi-thread CPU guard technology,which is based on non-blocking broadcast response technology and multi-node collaboration technology,is proposed for Infrastructure as a Service(IaaS);A multi-NAS(NAS,Network Attached Storage)technology,three interface distributed parallel processing model and node-based passive load balancing scheduling strategy are worked out for Platform as a Service(PaaS).(2)A fast Digital Surface Model extraction service based on C-RSIP(C-DSM)is proposed to further explore the construction method and significance of cloud computing platform of remote sensing data processing.Three more methods,ETM,adaptive noise cancellation algorithm and hierarchical weighted grid fitting algorithm,have been introduced to improve the tradition terrain extraction technology.(3)A remote sensing data processing system name as APGS based on C-RSIP has been developed and equipped in the China Centre for Resources Satellite Data and Application(CRESDA)and TH Satellite Ground Data Handling Centre.The APGS system architecture,system environment,system deployment and system implementation are discussed in detail,and the main function modules and some experimental results are presented in the following sections.The actual operation of the APGS shows that the C-RSIP model is reasonable in structure and reliable in performance and the improved fast terrain extraction algorithm is remarkable.The experimental results show that C-RSIP can effectively enhance the large remote sensing data processing capabilities,and process the data of satellite TH-1 and satellite ZY-3 into scenes,Digital Elevation Models(DEMs)and Digital Orthophoto Maps(DOMs)in real-time(that is the data is obtained and produced on the same day),the accuracy of which meets the requirement of 1:50000 scale.The proposed fast terrain extraction algorithm is more efficient than traditional methods,e.g.the ETM algorithm is nearly 20 times faster than the Semi-Global Matching(SGM)algorithm,and the hierarchical weighted grid fitting algorithm has no limit on the number of discrete points.Furthermore,APGS provides a new way for other industries to build cloud computing platform.