Dynamic motion-blur suppression and super-resolution localization of UAV targets using single-photon LiDAR under ultra-low SNR

Long-range detection and localization of UAVs remain extremely challenging, as motion-induced spatiotemporal distortion and weak echo returns severely degrade imaging quality. Although single-photon LiDAR offers photon-level sensitivity and long-range capability, its performance in dynamic UAV detection under low SNR remains constrained by motion blur, sparse photon measurements, and detector resolution, leading to unreliable target detection and localization. We present a five-dimensional motion-aware reconstruction (5D-MR) framework that jointly models the temporal (time-of-flight), spatial (row and column position), photon-count, and frame-sequence features of triggered photon events to perform signal preprocessing and dynamic target detection. Signal photons are robustly extracted from an overwhelming background via a two-stage RANSAC temporal fitting followed by DBSCAN clustering in the frame domain. Trajectory prediction combined with subpixel interpolation yields motion compensation and super-resolved localization beyond the sensor’s native resolution. Simulation results demonstrate that, even at an SNR of −9.03dB with 0.5343 photons per pixel, dynamic targets are localized with angular errors below 0.1 pixel and range errors within 0.5 m. Field experiments further validate the method for a UAV target at 1.8 km, achieving motion-blur suppression and super-resolution location reconstruction at an SNR of −8.97dB with only 0.3550 photons per pixel, resulting in angular precision better than 0.01 ∘ and size estimation errors below 0.02 m, representing nearly two orders of magnitude improvement in resolvable localization accuracy. The 5D-MR framework enables accurate 3D localization of dynamic UAVs under extreme photon-starved conditions, surpassing detector-array resolution limits and extending single-photon LiDAR to remote airspace monitoring and ultra-low-SNR dynamic imaging.