散乱イメージングのためのマルチチャンネル高精度FPGA組み込み時間相関単一光子計数システム

Scattering imaging techniques, demonstrated using time-correlated single-photon counting (TCSPC) systems, have significant industrial applications in areas such as remote sensing, robotic vision, and autonomous driving. This paper presents a multi-channel, high-linearity TCSPC system based on a Xilinx UltraScale (XC7Z100-2FFG9001) field-programmable gate array (FPGA). The proposed TCSPC system achieves a temporal resolution of 10 ps and exhibits differential nonlinearity within the range of [-10, 9.9] ps and integral nonlinearity within [-14, 9.9] ps. To demonstrate the capabilities of the proposed TCSPC system, we conducted confocal imaging experiments to visualise objects obscured by scattering media. We introduce what we believe to be a novel image reconstruction method based on the plug-and-play (PnP) framework, which leverages fast and flexible denoising neural networks: FFDNet for 2D imaging and FastDVDnet for 3D imaging. By integrating the scattering forward model with these denoising networks, we implement a high-performance reconstruction approach that enables high-quality imaging of complex scenes. Both simulations and experiments were conducted to validate the superior performance of our proposed method.