为了达到傅里叶变换成像光谱仪(FTIS)数据快速重建的目的,使用GPU并行计算技术设计了基于CUDA(compute unified device architecture)的成像光谱仪快速数据重建优化算法。采用CUDA下的CUFFT库和CUDA并行计算内核,以达到加快成像光谱仪快速数据重建。结果表明,基于CUDA的并行计算技术能有效调动GPU的硬件资源,可大幅度提高光谱重建处理任务的计算效率。如果将该技术应用到更多核的并行计算工作站上,那么单台计算机完成干涉成像光谱仪数据的实时处理任务将成为可能。 In order to achieve the rapid reconstruction of FTIR data, GPU-based parallel computing technology is used to design a fast data reconstruction algorithm based on CUDA (imaging unified device architecture). Using CUDA CUFFT library and CUDA parallel computing kernel, in order to accelerate the rapid reconstruction of imaging spectrometer data. The results show that the parallel computing technology based on CUDA can effectively mobilize the GPU hardware resources and greatly improve the computational efficiency of spectral reconstruction processing tasks. If this technology is applied to more nuclear parallel computing workstations, then a single computer to complete the interference imaging spectrometer real-time data processing tasks will be possible.