Abstract

Single-photon LiDAR offers photon-level sensitivity for long-range detection, yet its application to dynamic targets like low-altitude Unmanned Aerial Vehicle (UAV) is bottlenecked by motion-induced temporal-averaging bias. When target motion exceeds spatial resolution within a sampling interval, conventional multi-frame accumulation blurs the instantaneous spatial and temporal coordinates. Here, we introduce a photon-aware three-dimensional dynamic tracking framework (P3DTT) that suppresses this bias by transitioning to a multi-dimensional, photon-event-driven observation space.

By integrating spatio-temporal-frame photon-event statistical modeling with a Random Sample Consensus-based iterative fitting scheme, P3DTT extracts instantaneous, unbiased target positions under extreme photon starvation. Monte Carlo simulations demonstrate accurate 3D-position correction and tracking of target moving up to 188.8 m/s at 2 km under 0.2184 photons per pixel and a Signal-to-Noise Ratio (SNR) of −8.37 dB. The photon-arrival time fitting error is reduced from 4.71 ns to 0.33 ns (about 14-fold improvement), while the 3D tracking error reached 0.16 m. Field experiments conducted with the custom-developed Geiger-mode avalanche photodiode (Gm-APD) single-photon LiDAR demonstrate reliable UAV tracking at 2.5 km under only 0.2731 photons per pixel and an SNR of −7.68 dB. P3DTT consistently recovered unbiased instantaneous 3D-position, achieving tracking errors below 0.43 m.

This work establishes a new computational paradigm, pushing single-photon LiDAR beyond integral imaging towards instantaneous perception for low altitude monitoring.