基于SVM的烟叶光谱分级

Infrared Spectrum of Tobacco Grading based on Support Vector Machine

文中将基于统计学理论的支持向量机SVM(Support Vector Machine)与红外光谱分析技术结合,以50个烟叶样本作为实验材料,对两类烟叶进行分级。为了获得更好的定性分析结果并且简化网络输入维数,首先利用小波压缩对复杂光谱数据进行预处理。然后通过SVM建立烟叶分级模型。实验中采用高斯径向基函数(RBF)为核函数,根据SVM的不同输入量调整核参数建立最佳SVM模型,实验表明:对训练样本的正确识别率为100%,测试样本正确识别率为93.10%。

Support vector machine（SVM）based on the statistical theory and infrared spectrum analysis are integrated to classify the tobacco leaves, and in this classification, 50 tobacco leaves are employed as the test samples. For the purpose to obtain better results of qualitative analysis and reduce the input dimension of the network, the wavelet compression is adopted to pretreat the complicated spectra data and establish the tobacco classification model though SVM. The radial basis function is adopted as a kernel function of SVM, and the effect of RBF parameter, which could be adjusted according to the different input of SVM, is also investigated. The training set composed of 23 samples and the testing set composed of 27samples. The correct classification ratio of the training set is up to 100%, while that of the testing set is up to 93.10%.