Single-photon lidar: statistical analysis, high-flux imaging, and velocimetry
OA Version
Citation
Abstract
Single-photon lidar (SPL) systems leverage the sensitivity of single-photon avalanche diodes (SPADs) to achieve long-range, high-resolution 3D imaging. Despite these strengths, the technology faces many critical challenges that limit its real-world deployment. This thesis addresses three limitations---the reliance on knowing the ambient background flux, performance degradation in high-flux environments due to SPAD dead time, and the inability to measure velocity---by developing probabilistic models and corresponding estimators. First, we introduce a joint maximum likelihood (ML) estimator that simultaneously determines signal flux, background flux, and depth, enabling real-time adaptation to varying ambient light without requiring prior flux knowledge. Second, we develop computationally efficient joint ML estimators for high-flux operation under SPAD dead time. Simulations and experiments show the clear superiority of the free-running mode, in which the SPAD is reactivated immediately after the dead time, over the conventional synchronous mode, in which reactivation occurs at the beginning of the next repetition period. The free-running mode is robustly depth-independent, outperforms the synchronous mode even under extreme dead-time conditions, and has a higher optimal flux than the conventional ``5% rule.''For instance, at a signal-to-background ratio of 0.1, the optimal flux is approximately eight photons per repetition period, ie, ~ 800%. Finally, we pioneer Doppler SPL, a methodology that integrates target velocity into the probabilistic model as a Doppler shift in the pulsing repetition period. By employing Fourier analysis of Poisson processes and ML estimation, this approach jointly estimates velocity alongside all other scene parameters from absolute detection times. Collectively, these contributions significantly expand the operational envelope of SPL, transforming it into a robust, high-performance sensing modality capable of handling dynamic backgrounds, mitigating dead-time effects, and simultaneously measuring distance, velocity, and reflectivity.
Description
2026
License
Attribution 4.0 International