A New Multi-Channel Triangular FMCW LADAR Signals Denoising Method Based on Improved SMVMD

Abstract

Due to the short coherent integration time and other issues, the echo signal is seriously contaminated by noise, which reduces the target recognition accuracy of frequency modulation continuous wave (FMCW) laser radar (LADAR) in three-dimensional imaging. To solve it, this paper proposes a denoising method combining the improved dung beetle optimizer (DBO), successive multivariate variational mode decomposition (SMVMD), and singular-value decomposition (SVD). In our method, the improved DBO is applied to find the optimal balance parameter for decomposition; SMVMD jointly and adaptively decomposes multi-channel signals into intrinsic mode functions (IMFs) with aligned center frequencies and finds the target IMF with the optimal peak side lobe ratio (PSLR) among all decomposition results. To find possible multi-target peaks, the maximum singular value in the SVD of the target IMF is used as a threshold to filter the singular values in each IMF. The denoised signal can be obtained by accumulating the reconstructed IMFs with the low-rank approximation method. Finally, the targets are filtered by the frequency differences between the pulse pressure peaks of the opposite frequency-modulated signals from the same period. The proposed method can suppress more noise and extract appropriate target peaks for signals that are indistinguishable to peaks by amplitude, which is verified using actual FMCW LADAR three-dimensional imaging data.