Parallel Spectrum Reconstruction in Fourier Transform Imaging Spectroscopy Based on the Embedded System

摘要

In this paper, we design a parallel-processing pipeline for spectrum reconstruction in Fourier transform imaging spectroscopy (FTIS), which works well with the embedded system, NVIDIA Jetson TX2. This embedded system has great performance in parallel computing and can be developed easily by programmers using CUDA C in a single development board. This is very important for data processing on satellite and mobile devices. On the other hand, because of the huge amount of interference data acquired by the Fourier transform spectrometer, traditional interference data processing mechanism is not efficient and time-saving. These data should be processed in a fast way for real time, especially on satellite, to save memory and bandwidth. We take advantage of parallel computing to enable higher efficiency and reduced operation time. Furthermore, traditional serial algorithms for processing interferograms on the ARMs are introduced for comparison. The experimental results show that our parallel spectrum reconstruction pipeline has much higher performance than the serial one, and for huge data, our parallel mechanism also achieves great result in high performance.