基于无信息变量消除法和连续投影算法的可见-近红外光谱技术白虾种分类方法研究

应用无信息变量消除法结合连续投影算法对可见-近红外光谱区进行有效波长的选择,选择后的波长作为输入变量建立最小二乘-支持向量机模型,对白虾属中三种典型种,脊尾白虾、秀丽白虾和东方白虾进行鉴别分类.实验采用Kennard-Stone算法选取150个样本作为建模集,50个样本作为预测集,通过UVE-SPA优选了数值分别为392、431、517、551、595、627、676、734、760、861、943和1018 nm的12个波长为LS-SVM的输入变量,建立了白虾种分类模型.该模型对50个预测集样本检验的准确率达到了92.00%.结果表明,采用可见-近红外光谱对白虾种进行鉴别是可行的,UVE-SPA能够有效地进行波长选择,使LS-SVM模型获得最优的分类结果.

CLASSIFYING THE SPECIES OF EXOPALAEMON BY USING VISIBLE AND NEAR INFRARED SPECTRA WITH UNINFORMATIVE VARIABLE ELIMINATION AND SUCCESSIVE PROJECTIONS ALGORITHM

Using visible-near infrared spectra to classify different species of exopalaemon was studied. Successive projections algorithm (SPA) combined with uninformative variable elimination (UVE) were used to select effective wavelengths from visible and near infrared (Vis-NIR) bands. The selected effective wavelengths were set as inputs of least square-support vector machine (LS-SVM) for the classification of three typical exopalaemon species, namely, E. carincauda, E. modestus and E. orientis. Kennard-Stone algorithm was used to select 150 samples for calibration and the remaining 50 samples for prediction. Twelve effective wavelengths were selected by UVE-SPA, and they were 392,431,517, 551,595, 627, 676, 734, 760, 861, 943 and 1018 nm. The correct rate is 92.00% for classifying samples in prediction set by LS-SVM model based on these twelve effective wavelengths. The overall results demonstrate that it is feasible to utilize Vis-NIR spectroscopy to classify different species of exopalaemon, and UVE-SPA can extract the most effective wavelengths to build the LS-SVM model with an optimal classification result.