为了实现木器漆的快速无损检测以及精确识别与分类,采集并获取了晨阳等3种品牌木器漆样本的拉曼光谱,并考察了基线校正、Savitzky-Golay九点平滑法、一阶导数和二阶导数等不同预处理方法的处理效果,建立了特征波段比值、Fisher判别、K近邻(KNN)模型。结果表明:特征波段比值法能以1358 cm -1/1239 cm -1表征3种木器漆的特征;基于Fisher判别的基线校正、平滑和二阶导数处理的拉曼光谱模型的分类准确率最高,能实现100%区分;在相同的预处理下,KNN判别模型的准确率仅为88.5%。基于二阶导数的拉曼光谱结合特征波段-Fisher-KNN法能为不同品牌木器漆的准确检测提供一种新的快速无损分析手段,具有普适性和一定的借鉴意义。

In this study, the Raman spectra of three brands of wood lacquer samples, such as Chenyang brand, are examined to realize the rapid nondestructive detection as well as the accurate identification and classification of wood lacquer. Subsequently, the processing effects of different preprocessing methods, such as the baseline correction, Savitzky-Golay nine-point smoothing, first derivative, and second derivative, are investigated. Three models of characteristic-band ratio, Fisher discriminant, and K nearest neighbor (KNN) are also established. The experimental results indicate that the characteristic-band ratio method is capable of characterizing the features of the three wood lacquer samples with 1358 cm ^-1/1239 cm ^-1; further, the Raman spectral model combined with Fisher discriminant based preprocessing methods of baseline correction, smoothing, and second derivative exhibits an optimal classification accuracy of 100%. However, the accuracy of the KNN discriminant model is limited to 88.5% under the same preprocessing. Therefore, the second-derivative based Raman spectroscopy combined with the characteristic-band-Fisher-KNN method can provide a new rapid nondestructive analysis method for the accurate detection of different brands of wood lacquer, exhibiting universality and certain reference significance.