A High-Precision LiDAR Full-Waveform Echo Decomposition Method for Parameter Extraction

Aliased echo signal decomposition are the core of LiDAR systems. The multi-component overlap and non-Gaussian components of echoes pose challenges to the accurate extraction of waveform parameters. We proposed a convolutional decomposition method combining Gaussian and skew-normal models (CDGSN) to adapt to the echoes of different LiDAR systems and different scatterers. Using a layered extraction process (LEP) to effectively identify overlapping components. We compared the CDGSN method with four decomposition methods (Gaussian, B-spline-based, multi-Gaussian, and DRET) on synthetic echoes with known parameters and Global Ecosystem Dynamics Investigation (GEDI) satellite LiDAR data. Results indicate that CDGSN method achieves superior component detection, waveform fitting, parameter extraction, and ranging accuracy. It can obtain reliable target detection rate and parameter extraction accuracy in urban areas containing buildings and trees. Therefore, the CDGSN method offers computational efficiency, model flexibility, and strong potential for diverse non-Gaussian waveforms.