Research on Denoising of UV-Vis Spectral Data for Water Quality Detection with Compressed Sensing Theory Based on Wavelet Transform
ZHAO Ming-fu1, 2, TANG Ping1, 2, TANG Bin1, 2, 3*, HE Peng3, XU Yang-fei1, 2, DENG Si-xing1, 2, SHI Sheng-hui1, 2
1. Key Laboratory of Modern Optoelectronic Detection Technology and Instrument, Chongqing University of Technology, Chongqing 400054, China
2. Chongqing Key Laboratory of Optical Fiber Sensing and Photoelectric Detection, Chongqing University of Technology, Chongqing 400054, China
3. Key Laboratory of Optoelectronic Technology and Systems of Ministry of Education, Chongqing University, Chongqing 400044, China
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 UV-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-threshold 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.
赵明富, 唐 平, 汤 斌, 何 鹏,徐杨非, 邓思兴, 石胜辉. 基于小波变换的压缩感知理论对水质检测紫外-可见光谱数据的去噪研究[J]. 光谱学与光谱分析, 2018, 38(03): 844-850.
ZHAO Ming-fu, TANG Ping, TANG Bin, HE Peng, XU Yang-fei, DENG Si-xing, SHI Sheng-hui. Research on Denoising of UV-Vis Spectral Data for Water Quality Detection with Compressed Sensing Theory Based on Wavelet Transform. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2018, 38(03): 844-850.
[1] TANG Bin, WEI Biao, WU De-cao(汤 斌, 魏 彪, 吴德操). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2014,34(11): 3020.
[2] ZHOU Shi-long, GONG De-ren, DUAN Deng-ping(周世龙, 龚德仁, 段登平). Sensor and Micro Systems(传感器与微系统), 2015, 34(3): 154.
[3] TANG Bin, WEI Biao, MAO Ben-jiang(汤 斌, 魏 彪, 毛本将). Progress in Lasers and Optoelectronics(激光与光电子学进展), 2014(4): 197.
[4] Feng Z, Liang M, Chu F. Mechanical Systems and Signal Processing, 2013, 38(1): 165.
[5] Al-Qazzaz N K, Ali S, Ahmad S A, et al. Biomedical Engineering and Sciences (IECBES), 2014 IEEE Conference on. IEEE, 2014. 214.
[6] Zhu W, Chen B X. Multidimensional Systems and Signal Processing, 2015, 26(1): 113.
[7] CHEN Yu-shan, ZHANG Xiong-wei, YANG Ji-bin(陈栩杉, 张雄伟, 杨吉斌). Acta Automatica Sinica(自动化学报), 2015, 42(3): 335.
[8] Feng W, Fengwei C, Jia W. Open Electrical & Electronic Engineering Journal, 2015, 9: 74.
[9] ZHANG Jia-yan, HE Wei-ji, CHEN Qian(庄佳衍, 何伟基, 陈 钱). Chinese Journal of Optics(光子学报), 2015, 44(12): 70.
[10] TIAN Wen-biao, KANG Jian, ZHANG Yang(田文飚, 康 健, 张 洋). Electronic Journal(电子学报), 2014, 42(6): 1061.
[11] LI Hui-juan, GONG Xian-yong, LI Ying-cheng(李会娟, 巩现勇, 李英成). Science of Surveying and Mapping(测绘科学), 2014, 39(4): 131.
[12] Metzler C A, Maleki A, Baraniuk R G. IEEE Transactions on Information Theory, 2016, 62(9): 5117.
[13] Wang R, Yang Z, Liu L, et al. ACM Transactions on Graphics (TOG), 2014, 33(2): 18.
[14] Lopes M E. IEEE Transactions on Information Theory, 2016, 62(9): 5145.
[15] Majumdar A, Ansari N, Aggarwal H, et al. Signal Processing, 2016, 119: 136.
[16] WANG Qiang, LI Jia, SHEN Yi(王 强, 李 佳, 沈 毅). Chinese Journal of Electronics(电子学报), 2013, 41(10): 2041.
[17] Ma Q, Quan X, Zhong Y, et al. Cogent Engineering, 2016, 3(1): 1247611.
[18] YUAN Qin, WU Xuan-gou, XIONG Yan(袁 琴, 吴宣够, 熊 焰). Computer Science(计算机科学), 2014, 41(3): 314.