论文部分内容阅读
针对微分法在有效消除光谱背景和基线漂移的同时会增加光谱噪声的问题,把最新发展的经验模态分解方法(EMD)引入到近红外光谱处理中来,以烟草的一阶导数近红外(NIR)光谱为研究对象,探讨经验模态分解在近红外光谱预处理中的应用,并与小波变换消噪效果进行了对比分析。结果表明,用基于经验模态分解去噪后的光谱进行分析,预测集的决定系数r2由去噪前的0.9705提高到0.9832,均方根误差(RMSEP)由去噪前的0.5606降为0.3310,比基于小波变换的分析结果略高。因此,经验模态分解方法对消除光谱的噪声是有效的,有效地提高了光谱的分析精度和模型的稳定,为近红外光谱预处理提供了一种新方法。
Aiming at the problem that the differential method increases the spectral noise while effectively eliminating the spectral background and the baseline drift, the latest developed empirical mode decomposition method (EMD) is introduced into the near infrared spectroscopy processing. The first derivative of the near infrared NIR) spectrum as the research object, the application of empirical mode decomposition in NIR pretreatment is discussed and compared with the wavelet transform denoising effect. The results show that the RMS r of the prediction set is improved from 0.9705 before the denoising to 0.9832 and the root mean square error (RMSEP) is reduced from 0.5606 before denoising to 0.3310 by analyzing the spectrum after denoising based on empirical mode decomposition. Slightly higher than the result of wavelet transform based analysis. Therefore, the empirical mode decomposition method is effective to eliminate the spectral noise, effectively improves the spectral analysis accuracy and the stability of the model, and provides a new method for NIR pretreatment.