基于傅里叶变换近红外和支持向量机的霉变玉米检测

Detection of Moldy Corns with FT- NIR Spectroscopy Based on SVM

利用傅里叶变换近红外光谱对霉变玉米进行检测。运用波数范围在12 000～4 000 cm^-1的FTNIR系统进行不同霉变程度样品光谱信息的采集,并利用支持向量机（Support Vector Machine,SVM）进行数据分析。结果显示,利用原光谱数据,以主成分分析（PCA）提取的前5个主成分作为输入,选用径向基函数（RBF）作为SVM核函数,并运用网格划分寻优法寻得的最优参数C、γ,所建立的分类模型最佳。SVM分类模型对训练集和测试集的预测准确率分别达到93.3%和91.7%,对独立样品集的预测准确率达到87.8%,表明基于FT-NIR和SVM进行霉变玉米的检测是可行的。

The moldy corns have been detected by the Fourier Transform Near Infrared （ FT - NIR） Spectroscopy. A FT - NIR spectrometer of spectral range between 12 000 cm^-1 to 4 000 cm^-1 was used to acquire spectral data of every kernel. The data was analyzed by Support Vector Machine （SVM）. Principal component analysis （PCA） was used to reduce the dimensions of the original spectral data. The experimental results showed that when the five principal components （PCs） were selected as the input of SVM, and the radial basic function （RBF） was used as the kernel function, as well as the best parameters C and Y were selected by grid search,, the established SVM classifica- tion model was optimal, and the classification accuracies of 93.3% , 91.7% and 87.8% respectively for the training set, testing set and independent test set were available for being achieved. The result indicated that it was feasible to identify and classify different degree of moldy corn grain kernels by FT - NIR and SVM.