Optimalization-based depth estimation method for single-photon LiDAR

The pile-up effect inherent in the common synchronous gating mode of single-photon avalanche diodes (SPADs) limits detection accuracy under high photon flux. Traditional correction methods based on recursion insufficiently utilize full time-domain information, leading to significant errors under moderate to high noise intensity. To address this, this paper constructs a smooth photon detection probability distribution model for complex illumination conditions, based on the Gaussian waveform prior of the signal light and the Poisson distributed response model of the detector. Leveraging the differentiability of this model, we propose an iterative high-precision imaging parameter inversion algorithm by introducing adaptive moment estimation (Adam) and an adaptive parameter initialization mechanism. Extensive simulations and real-system testing demonstrate that the proposed method exhibits stronger noise robustness compared to existing algorithms. Specifically, under the conditions with the signal photon count of 0.1 and the noise photon count of 0.1, the proposed method reduces the root mean square error (RMSE) by 5.71 cm (from 12.766 to 7.056 cm).