宽波段微型光谱仪的小波奇异值差分去噪

Wavelet singular value difference de-noising for broadband micro spectrometer

为克服传统光谱仪器价格昂贵、体型庞大等缺点,设计了一款宽波段微型光谱仪,测量波段为200~1100nm,光谱的像元分辨率均在0.5 nm以内,具有较好的分光特性。为提高光谱仪的测量精度、降低光谱仪硬件和光学系统引入的噪声,提出了一种小波奇异值差分算法,使用小波阈值去噪方法对高频信号的低层小波系数进行消噪处理,对高频信号的全部小波系数进行奇异值分解,并采用差分法选择奇异值阶次。结果显示,信噪比可达到20.9600,去噪效果较好;均方根误差值为1510.3,波形相似系数值为0.9960,去噪较彻底且能保留光谱有效特征,对宽波段微型光谱仪的去噪具有很好的可行性。

In order to overcome the shortcomings of traditional spectroscopic instruments,such as high price and large size,a broadband micro spectrometer was designed and developed.The spectral resolution of the spectrum was 200-1100 nm,and the pixel resolution of the spectrum was all less than 0.5 nm,so it had good spectral characteristics.In order to improve the measurement accuracy of the spectrometer and reduce the noise introduced by the spectrometer hardware and optical system,a wavelet singular value difference algorithm is proposed to deal with the spectral signal.The wavelet threshold is used to de-noise the low-level wavelet coefficients of the high-frequency signal,and the singular value is used to decompose its whole wavelet coefficients.The singular value de-noising order is selected by the difference method.The results show that the signal noise ratio(SNR)value can reach 20.9600,so the de-noising effect is good;the root mean square error(RMSE)value is 1510.3,the normalization cross correlation(NCC)value is 0.9960,the de-noising is complete and the spectral effective feature can be preserved.The proposed algorithm has a great feasibility for the de-noising of the broadband micro spectrometer.