高分辨率傅里叶变换成像光谱仪光谱重建并行算法

Spectral reconstruction parallel algorithm for high resolution Fourier transform imaging spectrometer

高分辨率傅里叶变换成像光谱仪具有高空间分辨率和高光谱分辨率的特点,但光谱重建时间冗长。通过对傅里叶变换光谱重建流程分析,为研制的1024pixel(光谱维)×1024piexl(像宽)×1024piexl(像高)高分辨率紫外傅里叶变换成像光谱仪的数据立方体反演,设计了一种并行优化算法。实验表明,在6核处理器上对512M和2G的数据立方体进行变换,时间分别只需88.33s和489.75s,加速比分别为3.70和3.04,大幅度提高了运算效率。如将该算法应用到更多内核处理器上,可得到更高的加速比和更少的运算时间。

High resolution Fourier transform imaging spectrometer has both high spatial resolution and high spectral resolution features, but the spectral reconstruction is time consuming. For the inversion of data cube of self-developed 1024 pixel （spectral dimension）×1024 pixel （image wide） × 1024 pixel （image high） high resolution ultraviolet Fourier transform imaging spectrometer, a parallel optimization algorithm is designed by analyzing the process of Fourier transform spectrum reconstruction. The experimental results show that the transforming time of 512M and 2G data cube are only 88. 33 seconds and 489. 75 seconds on the 6 core processors. Speedup ratio is about 3. 70 and 3.04 respectively, which improve the operation efficiency effectively. If the algorithm is applied to more cores processor, higher speedup ratio and less computation time can be obtained.