Development of GPU-Accelerated Pre-processing Chain for LAPAN-A2 Multispectral Imagery

This paper reported the development of a newly accelerated chain for pre-processing raw data acquired by LAPAN-A2 multispectral sensor. The acceleration schema was achieved by making full use of Graphics Processing Unit (GPU) to handle parallel processing in raw data encoding, image demosaicing, and radiometric-geometric correction stages. Based on the experiment conducted on NVIDIA GeForce GT 1030 GPU card, the developed chain was found capable of resulting level-1 imagery accurately of up to 3.72 times faster than that of similar implementation on Core-i7 8700 CPU. It was also found that the GPU device boosted the demosaicing stage more than 10 times faster. The result of this research demonstrates the utility of the attached multiprocessors of GPU to reduce the execution time of satellite imagery pre-processing.