Research On Remote Sensing Image Real-time Processing Method Based On Stream Model

In Recent year,Satellite remote sensing for Earth observation technology has entered a new era of diversified data acquisition and quantitative information,and many special fields require real-time processing character.The traditional satellite data processing faces major challenges,such as data storageand data management in the field of emergency(mobile target detection etc.).And remote sensing image processing must overcome the limitation of processing environment and resource,and relalize high performance fast real-time processing under mobile environment.A number of recent studies have shown that conventional satellite data ground processing systems and algorithms do not take into account the computational efficiency and volumetric power consumption.Thus the potential satellite processing system must obscure the stand-alone manual processing mode.At present,the level of hardware and software has been greatly improved,and high performance processing technologies such as cluster computing have been widely used.The phenomenon of high-performance computing technology,including remote sensing and mapping,has become very common.Therefore,how to reconstruct the optical satellite remote sensing pre-processing algorithm by using high-performance technology in the actual situation of the massive rapid influx of remote sensing data,and finally construct a dynamic and balanced parallel computing frame to solve the problem of high-performance remote sensing image preprocessing.Based on the real-time processing of optical satellite data and the core algorithm of GPU parallel program foundation,the paper designs a remote sensing image stream processing architecture.This architecture proposes an improved consistency hash algorithm as a strategy for dynamic load regulation of nodes.For the real-time computing requirements,the paper adopts two model:distributed processing model and pipeline processing model.The architecture uses a multi-tier storage architecture for I/O optimization and Intel C ++ Compiler compiler for compiler optimization,for in-depth software optimization.The flow framework proposed in this paper is based on the streaming processing mode,and realizes the real-time fast processing of the remote sensing image in the emergency field.The paper do the comparative test on the storage architecture,compiler optimization,GPU algorithm optimization,which between the stream architecture algorithm and the traditional algorithm,which data areThe ZY-3 etc.domestic optical satellite data.The test results show that the multi-level storage architecture constructed in this paper has 123 times in reading and writing in the ramdisk than in the HDD,which significantly improves the efficiency of file transfer.The compiler optimization strategy shortens the running time of 1/3 of the software and accelerates the CPU clock speed;On the whole,the algorithm of full-color CCD orthophoto image generation time is reduced to 33s or less.