Enhanced Doppler tomography LiDAR imaging using the MACE framework and deep denoisers

Doppler tomographic LiDAR (DTL) is a groundbreaking technique for long-range imaging of rapidly rotating targets, offering significant advantages in applications such as drone surveillance, orbital object tracking, and space debris characterization. However, its imaging performance and operational range are severely limited by inherent shot noise and non-stationary speckle artifacts. To overcome these challenges, we propose what we believe to be a novel DTL reconstruction framework based on multi-agent consensus equilibrium (MACE), which synergistically integrates a refined physical imaging model with advanced data-driven denoising. Our approach introduces three key innovations: (1) A Bayesian-optimized physical model that rigorously characterizes noise propagation dynamics; (2) Two specialized deep denoisers explicitly designed to address the distinct statistical properties of shot noise and speckle noise; (3) A seamless MACE-based integration that harmonizes the physical model and denoisers, enabling robust and adaptive noise suppression across diverse environments. Extensive simulations and real-world experiments demonstrate that our method significantly outperforms state-of-the-art techniques, especially under low signal-to-noise ratio conditions. In field tests, we successfully imaged a target rotating at 2,400 RPM from a 40-meter standoff distance with a signal-to-noise ratio as low as -10 dB, where competing methods failed to recover structural details. This work establishes a flexible and robust framework for high-resolution imaging of fast-rotating targets in realistic noise environments, advancing the practical viability of DTL for critical aerospace applications.