Speckle suppression for UAV tracking using single-photon LiDAR with vibrating MMF and ST-DKDE

The widespread deployment of low-altitude, slow-speed, and small-sized (LSS) UAVs poses significant challenges to airspace security monitoring, where achieving both high-precision detection and long-range coverage remains difficult. Gm-APD-based single-photon LiDAR systems greatly enhance detection sensitivity and range, but the use of multimode fiber (MMF) introduces strong speckle noise, resulting in non-uniform echo distributions, model mismatch, and potential tracking failure. To address this issue, we propose a transmitter-side fiber vibration strategy to physically suppress speckle fluctuations. Building on this, we develop a spatio-temporal joint dynamic kernel density estimation (ST-DKDE) reconstruction algorithm, enabling robust photon-intensity recovery. This enables robust intensity recovery, which in turn supports high-precision tracking of aerial targets using an improved MeanShift-Kalman algorithm. Monte Carlo simulation verification shows that the algorithm can reduce the average speckle contrast from 1.1999 to 0.1039 (a decrease of over 90%). The proposed algorithm was validated through field experiments using a Gm-APD single-photon LiDAR system for UAV detection and tracking. In the first scenario, a UAV flying around the LiDAR's FoV at approximately 200 m under overcast conditions (visibility 4 km, ambient illumination 8,979 lux) showed an 83% reduction in speckle contrast, an average tracking accuracy of 0.89 cm, maximum error below 4.2 cm, and a 61.9% decrease in mean squared tracking error. In the second scenario, involving dynamic UAV flight from 300 m to 100 m with rapid turns and close-range approaches under clear conditions (visibility 10 km, illumination 55,120 lux), the proposed method achieved up to 82.6% reduction in mean tracking error and 88.0% reduction in RMSE compared to traditional peak-based methods, maintaining tracking errors consistently below 3 pixels. These results demonstrate that effective speckle suppression combined with spatio-temporal modeling enhances inter-frame matching and tracking stability, providing a robust solution for dynamic LSS UAV monitoring in complex airspace environments. This work establishes a solid foundation for high-precision, long-range LiDAR-based aerial surveillance.