对吉林一号光学卫星的应急快速处理方法研究

An Emergency Rapid Processing Method Based on JL-1 Optical Satellite

在CPU/GPU协同处理框架下,系统地探讨了光学遥感影像高性能处理方法。首先,实现光学卫星数据处理算法的GPU高效映射,将MTF补偿、传感器校正(包括波段配准和影像拼接)、系统几何校正等算法高效映射至GPU并行执行;在此基础上,为充分利用CPU高频化优势和加速I/O运算,利用Ramdisk技术在内存盘处理程序数据与结果输出,利用Intel C++Compiler编译优化框架。在GeForce GT 755M(GPU)、Mobile Quad Core Intel Core i7-4700MQ(CPU)的Windows系统环境下,对吉林一号卫星全色影像和多光谱影像原始数据进行0~2级产品的光学遥感影像预处理。实验结果表明,与传统预处理算法相比,此高性能算法的处理时间缩短到了40s以下,最高加速比达到11.216,可满足对海量光学卫星遥感影像数据的应急快速处理需求。

A high-performance optical image processing method based on CPU/GPU collaborative processing frame work is put forward. First, we realized GPU effectively map of the optical satellite data processing algorithm, MTF com pensation, sensor correction（including band register and im- age stitching） and geometric correction were mapped to GPU for execution. To take full advantage of GPU high frequency advantage and accelerate I/O computation, Ramdisk tech- nique and Intel C＋ ＋ Compiler were designed to achieve the optimal performance. It is found that the processing time of this algorithm in Windows environment of GeForce GT 755M（GPU）, Mobile Quad Core Intel Core i7-4700MQ （CPU）, JL-1 satellite panchromatic image and multispectral image data were preprocessed. Experimental results show that, compared to traditional preprocessing algorithm, the processing time of our proposed algorithm is less than 40 s and the highest speedup ratio is 11. 216,which can meet the requirement of rapid emergency response to mass optical sat-ellite irrlage data.