摘要
针对直接采用Forman相位校正算法实现HJ-1A高光谱数据实时光谱复原处理系统过程中遇到的算法复杂、运算量大以及运算效率低的问题,提出了改进的Forman算法。该算法简化了相位数据的计算,并利用复数序列傅里叶变换的性质和实序列傅里叶变换的共轭对称性优化了计算过程。在分析改进前后算法关键步骤的浮点数运算量的基础上,以HJ-1A高光谱数据为实验对象,基于图形处理器(graphic processing unit,GPU)实现了改进的Forman算法。结果表明:在相同计算精度的情况下,改进的Forman算法运算速度提高了11.86%,为HJ-1A高光谱数据实时光谱复原处理系统的实现提供了行之有效的方法。
Problems such as complex algorithm,heavy computation and low efficiency occur when Forman method is directly used in the implementation of HJ-1A real-time hyperspectral data processing system.In this paper,an improved method is proposed.With this method,phase calculation is simplified and,at the same time,the calculation process is optimized by using the property of Fourier transform of the complex sequence and the conjugate symmetry of Fourier transform of real sequence.On the basis of analyzing the amount of floating point operations of the algorithm key step before and after the improvement,the improved method is eventually realized based on the graphic processing unit(GPU) as the experimental subject to HJ-1A hyperspectral data.The results show that the operating speed of the improved Forman algorithm is increased by 11.86% in the case of the same calculation accuracy.Thus,the improved Forman algorithm can provide an effective way to realize the HJ-1A real-time hyperspectral data processing system.
出处
《国土资源遥感》
CSCD
北大核心
2013年第3期97-101,共5页
Remote Sensing for Land & Resources
基金
"高分辨率对地观测系统"项目(编号:科工高分[2010]1481号)资助