High-fidelity ghost diffraction through complex media using a single-photon detector

Free-space optical (FSO) transmission in complex scenarios remains a challenge, especially at low-light levels. Here, we report a ghost diffraction system with a single-photon detector to enable effective and robust transmission through dynamic scattering media under photon-limited conditions. At the transmitter, each pixel of a signal is encoded into a two-dimensional (2D) random pattern via a single-layer convolutional neural network (SCNN). By using an all-ones matrix as an input and the random pattern as a convolution filter, SCNN can be designed to model the physical process of ghost diffraction and can scale the sum of each random pattern to be proportional to a corresponding pixel of the signal in an untrained manner. The generated 2D random patterns, serving as information carriers, are sequentially displayed in an FSO channel to modulate a laser beam. At the receiver, weak and scattered light intensities are detected by using a single-photon counting module. To verify the proposed ghost diffraction system, a series of optical experiments are conducted using varying water turbidities and different rotation speeds. Experimental results demonstrate that the proposed method can achieve high-fidelity and high-robustness FSO transmission in femtowatt-level low-light environments with random disturbances from dynamic and turbid water. The proposed ghost diffraction system with a single-photon detector offers a promising solution for high-fidelity FSO transmission in complex scenarios at low-light levels.