Research on Denoising of UV-Vis Spectral Data for Water Quality Detection with Compressed Sensing Theory Based on Wavelet Transform
Zhao Ming-fu,Tang Ping,Tang Bin,He Peng,Xu Yang-fei,Deng Si-xing,Shi Sheng-hui
DOI: https://doi.org/10.3964/j.issn.1000-0593(2018)03-0844-07
2018-01-01
Spectroscopy and spectral analysis
Abstract:It is of great significance to improve the measurement stability and accuracy of water quality detection system with direct spectrum method. Direct spectroscopy on-line water quality detection systems typically use long-lived, preheated pulsed xenon lamps and industrial-grade spectral detection devices for complex inspection environments. Since the whole spectral detection system is affected by the light source, the optical path and the photoelectric conversion device, the measured spectral data contains a large amount of noise, a wavelet denoising algorithm based on compressed sensing is proposed, which is compared with the traditional wavelet threshold denoising method. In this paper, the denoising was performed on the LTV-Vis spectra of the standard solution of potassium hydrogen phthalate with chemical oxygen demand of 200 mg . L-1. The compressed sensing algorithm is used to decompose the signal in the wavelet domain, and the high frequency coefficients are obtained. Using the random Gaussian matrix as the observation matrix of the compression sensing algorithm, the compression ratio is set to 2, and the high frequency coefficients are observed. The orthogonal matching algorithm is used to recover the sparsity of the high frequency wavelet coefficients to achieve the denoising. At the same time, for the traditional wavelet threshold denoising, the soft-thresh-old filtering method is used to denoise the spectral data, and the wavelet base is daubechies 4. In order to verify the feasibility of the noise reduction algorithm, the spectral signals of a stream and domestic sewage were collected, and the above two methods were used to denoise the spectral signal. The experimental results show that the compressed sensing algorithm based on wavelet transform is suitable for the on-line water quality detection system based on UV-Vis spectroscopy. The method can effectively denoise under the premise of preserving the absorption characteristics of the original spectral signal of the water sample, and the denoising effect is better than the wavelet threshold denoising algorithm. Compared with the wavelet threshold denoising algorithm, the SNR is increased by 12. 201 5 dB, the RMSE is neduced by 0. 009 3, and the PSNR is increased by 5. 299 dB. The proposed method not only avoids the problem of threshold selection in wavelet threshold denoising, but also effectively suppresses the noise in the reconstruction process. This method provides a new solution for direct spectroscopy to detect water quality parameters.