Wavelet and Empirical Mode Decomposition Denoising for GLAS Full Waveform Data

The full waveform data obtained by the geoscience laser altimeter system (GLAS) has different degrees of noise due to the system performance, the scattering of the propagation medium and the characteristics of the detection target, which affects the extraction of vertical structure parameters and the laser ranging accuracy of the GLAS. Therefore, the noise reduction effects of Gaussian filtering, wavelet threshold denoising and empirical mode decomposition (EMD) methods are analyzed and compared in this paper, and the performance of wavelet soft threshold denoising and wavelet improved threshold denoising methods as well as EMD-wavelet threshold denoising and EMD-Hurst noise reduction methods are further compared. The experimental results on six typical feature data show that the denoising results of wavelet threshold denoising and EMD-Hurst denoising methods are better than Gaussian filtering except EMD-wavelet threshold denoising. All three types of denoising methods, Gaussian filtering, wavelet threshold denoising and EMD denoising, have the best denoising effect on flat without slope. In addition, the wavelet improved threshold denoising method is better than the wavelet soft threshold denoising method, the signal-to-noise ratio (SNR) is increased by 10. 70%-45. 72% and root mean squared error (RMSE) is reduced by 32. 04%-81. 94%. The EMD-Hurst method is better than the EMD-wavelet threshold denoising method, the SNR is increased by 6. 38%-65. 70% and the RMSE is reduced by 13. 53%-33. 33%