Remote Sensing Data Real-time Processing Research Based On GPU

This paper mainly researches apaceborne synthetic aperture radar(SAR)imaging processing technology and the realization of its application in real-time GPU(graphic processor unit). The spaceborneSAR imaging processing requires massive computing and high-speed cache,the efficiency of traditiondal computing based on CPU processor is low.which is difficult to meet the demand for high-efficiency application of SAR imaging. Programmable graphic core processor uses multithreading technology, owns high computing power, it has a broad field of application prospects on high-performance general-purpose computing and high-efficiency SAR image processing.This paper describes the spaceborne synthetic aperture radar imaging principle,uses the Chirp Scaling algorithm which is easily implemented on hardware with simple structure, completes the implemention of synthetic aperture radar imaging processing based on the new computing platform CUDA and the use of graphiccore processor GPU with high-speed floatpoint-operation ability.This paper uses CUDA parallel computing architecture, preprocesses the blocks of image data through the creation of multiple threads;completes the track parameter settings and the calculationof doppler parameters by the co-processing of CPU and GPU;completes the phase factor calculation and multiplication of complex factor by multi-core parallel processing. Finally, this paper gives the experiment results and tests the performance, the analysis shows that the GPU-based image processing greatly enhanced imaging efficiency compared with CPU-based hardware solutions.