Robust UAV tracking under low-SNR via TRSCE-based speckle suppression in vibrating-MMF single-photon LiDAR

With the large-scale deployment of unmanned aerial vehicles (UAVs), long-range, high-precision detection and tracking face severe challenges. Although Gm-APD-based single-photon LiDAR systems exhibit outstanding sensitivity under ultra-low SNR and weak echo conditions, the employment of multimode fiber (MMF) coupling to achieve high energy output tends to introduce strong speckle noise, which significantly degrades echo photons’ stability and imaging quality, thereby reducing the accuracy of target tracking. This study proposes a speckle suppression method that integrates MMF vibration with the time-domain RANSAC with spatial consistency enhancement (TRSCE) reconstruction algorithm, which combines temporal RANSAC and spatial neighborhood structural consistency. Additionally, an improved MeanShift-Kalman algorithm is proposed to achieve robust tracking of UAVs under low SNR. Monte Carlo simulations demonstrate that under an average single-pixel echo photon count of 0.4481 and an SNR of −6.85 dB , the speckle contrast within the target region decreases significantly from 0.7592 to 0.0748, corresponding to a reduction of 73.2%. In field experiments conducted under weak solar illumination and light fog, the proposed Gm-APD single-photon LiDAR system demonstrated robust tracking capability of UAV targets at a distance of 200 m. At an SNR of approximately −4.53 dB , with an average signal return of 1.0859 photons per target pixel and a background noise level of 3.0829 photons, the system achieved a mean tracking error of 0.6572 pixels (approximately 6.44 cm) and an inter-frame model similarity of 0.9764. Compared to publicly reported speckle suppression methods, the proposed approach reduced the tracking error by approximately 45.3%. Furthermore, at a range of 300 m under strong solar illumination, with an average signal-photon count of 0.8769 and background noise of 4.0611 photons per pixel, the system achieved a mean tracking error of 0.6975 pixels (approximately 9.66 cm), improving accuracy by 43.2% compared to spatial-correlation-based algorithms. Finally, by integrating phased-array radar, the system extends effective detection and tracking distances of UAVs to 10 km, surpassing existing LiDAR capabilities. The proposed speckle suppression and robust tracking framework significantly enhances tracking stability of LiDAR operating in complex airspace environments and provides a viable theoretical foundation and engineering solution for future low-altitude defense and airspace security applications.