Adaptive Denoising for Airborne LiDAR Bathymetric Full Waveforms Using EMD-Based Multiresolution Analysis

It’s a key issue denoising airborne LiDAR bathymetric (ALB) full-waveforms in extracting underwater topography. Empirical mode decomposition (EMD) is suitable for the nonlinear and non-stationary characteristics of ALB full-waveforms and obviates the necessity for pre-set basis functions. Still, the direct discarding of noise-dominated Intrinsic Mode Functions (IMFs) by traditional EMD leads to over-smoothing or under-smoothing of signals. In this letter, An EMD-based multi-resolution analysis (EMD-MRA) method is suggested for ALB full-waveform denoising. This method utilizes the zero-padding technique based on the Discrete Cosine Transform (DCT) to achieve the second-layer decomposition of the noise-dominant IMFs obtained in the first-layer. Subsequently, Savitzky-Golay(S-G) filtering is employed for the noise-dominant IMFs from the second-layer decomposition. The intricate details embedded in the IMFs are meticulously extracted through a process of scaling the signal from a coarse to a fine resolution, ensuring the preservation of valuable information. The experiments, conducted using measurement data, demonstrate that the EMD-MRA method is effective in adaptively denoising the ALB full-waveform data, showcasing sufficient robustness. Compared with traditional EMD and wavelet threshold denoising (WTD) methods, the signal-to-noise ratio (SNR) of the denoising results using the proposed approach is improved by 6.769 dB and 0.971 dB in area a, and by 19.672 dB and 5.317 dB in area b. The root mean square error (RMSE) is reduced by 0.455 and 0.971 in area a, and by 0.979 and 0.47 in area b, effectively retaining the complex details of the original signal.