近红外光谱和多分类器融合的葡萄酒品种判别研究

发布时间：2018-12-29 20:12

【摘要】：将多种单分类器模型融合,并用融合后的模型对不同品种干红葡萄酒进行判别分析。用BRUKER MPA傅里叶变换型近红外光谱仪采集170个干红葡萄酒样品的近红外透射光谱,选取PLS-DA,SVM,Fisher和AdaBoost作为单分类器建模方法,分别建立葡萄酒品种判别模型,通过差异性度量值对单分类器进行筛选,得到差异性较大的四个单分类器作为基分类器,其中基分类器对测试集葡萄酒品种判别准确率最高为88.24%,最低为81.18%。然后通过加权投票机制对基分类器进行融合,融合后的模型对测试集葡萄酒品种判别准确率提高至92.94%,误判样品个数由单分类器最少的9个降为6个。实验结果表明多分类器融合所建立的模型优于传统近红外光谱定性分析一般采用单分类器模型结果,提高了葡萄酒品种判别的准确性,采用基于近红外光谱的多分类融合方法对葡萄酒种类判定具有可行性。
[Abstract]:The models of several single classifiers were fused and the different varieties of dry red wine were analyzed by using the fused models. The near infrared transmission spectra of 170 dry red wine samples were collected by BRUKER MPA Fourier transform near infrared spectrometer. PLS-DA,SVM,Fisher and AdaBoost were selected as the modeling method of single classifier. By screening the single classifier by the difference measure, four single classifiers with great difference were obtained as the base classifier, in which the accuracy rate of the base classifier for the wine varieties was 88.24 and 81.18 respectively. Then the base classifier was fused by weighted voting mechanism. The accuracy rate of wine variety discrimination in the test set was improved to 92.94 and the number of misjudged samples was reduced from 9 with the least single classifier to 6 by the fusion model. The experimental results show that the model established by multi-classifier fusion is better than that of the traditional NIR qualitative analysis by using single classifier model, which improves the accuracy of wine variety identification. The method of multi-classification fusion based on near-infrared spectrum is feasible for wine classification.
【作者单位】：中国农业大学信息与电气工程学院;中国农业大学食品科学与营养工程学院;
【基金】：国家自然科学基金项目(31101289) 智慧农业关键技术研究项目(15055340)资助
【分类号】：TS262.6;O657.33

【相似文献】