傅里叶变换红外光谱和紫外光谱数据融合对牛肝菌种类的鉴别

采集5?种共272?份牛肝菌样品的傅里叶变换红外光谱和紫外光谱，结合多光谱信息融合策略，建立牛肝菌种类快速鉴别的方法。多元散射校正（multiplicative signal correction，MSC）及二阶导数（second derivative，2D）等预处理方法对原始光谱进行优化，比较优化处理对区分不同种类牛肝菌影响；利用优化处理后的光谱数据及融合数据建立偏最小二乘判别分析（partial least squares discriminant analysis，PLS-DA）模型和支持向量机（support vector machine，SVM）判别模型。结果显示：1）经过2D和MSC预处理后，不同种类牛肝菌的PLS-DA鉴别效果优于未优化模型，表明2D＋MSC预处理优化了光谱信息并提高了分类准确度；2）基于傅里叶变换红外光谱、紫外光谱、低级融合和中级融合数据分别建立PLS-DA模型，预测正确率为86.87%、66.67%、78.89%和95.56%；建立SVM判别模型，预测正确率分别为88.89%、74.44%、91.11%和100.00%，表明中级融合技术对不同种类牛肝菌鉴别效果显著，优于其他技术；3）中级融合技术在PLS-DA模型和SVM判别模型中对样品的预测正确率分别为95.56%和100.00%，表明SVM判别模型对牛肝菌种类区分效果优于PLS-DA模型。采用中级融合技术建立SVM判别模型，快速鉴别牛肝菌种类，为牛肝菌种类鉴别和质量控制提供可靠、稳定的方法。

Data Fusion of Fourier Transform Infrared and UV Spectra for the Discrimination of Bolete Mushrooms

In this study, a rapid method using Fourier transform infrared (FTIR) and ultraviolet (UV) spectroscopies coupled with data fusion was established for the identification of five species of bolete mushrooms. The original spectra of 272 samples from different species were preprocessed and optimized by multiplicative signal correction (MSC) and second derivative (2D). The classification results from the original and the optimal data matrixes were compared. Then the single and fused spectral data were analyzed by partial least squares discriminant analysis (PLS-DA) and support vector machine (SVM). The results showed that 1) compared with the original spectra, the preprocessing methods of 2D and MSC could optimize the spectral information and improve the classification accuracy; 2) base on FTIR, UV, low-level data fusion and mid-level data fusion data matrixes, respectively, the prediction accuracy of the PLS-DA models was 86.87%, 66.67%, 78.89% and 95.56%, and the prediction accuracy of the SVM models was 88.89%, 74.44%, 91.11% and 100.00%, respectively, indicating that mid-level data fusion was better than the other methods; 3) in terms of the prediction accuracy, the classification ability of the SVM model with mid-level data was better than that of the PLS-DA model with mid-level data. In conclusion, the SVM model based on mid-level data fusion represents a rapid method for the effective identification of different species of bolete mushrooms. It can provide an effective method for identification and quality control of edible mushrooms.