基于多分类器融合的近红外光谱技术鉴别蜂蜜品种
张林1,韩美林1,杨琳1,王洋2
1. 商洛学院电子信息与电气工程学院,陕西商洛 726000; 2. 西北农林科技大学食品科学与工程学院,陕西杨凌 712100
Identification of honey varieties by near infrared spectroscopy
张林1,韩美林1,杨琳1,王洋2
1. School of Electromation Information and Electrical Engineering, Shangluo College, Shangluo 726000, China;
2. College of Food Science and Engineering, Northwest A&F University, Yangling 712100, China
摘要:建立基于多分类器融合的近红外光谱技术判别蜂蜜品种的方法。采用Fisher,SVM,PLS-DA 和AdaBoost
作为单分类器,分别建立蜂蜜品种的判别模型,通过差异性度量值分别对单个分类器进行筛选,得到差异性最大的3个分类器,将这3 个单分类器进行融合,将融合后的多分类器模型用于对蜂蜜品种的判别分析。单个分类器模型对蜂蜜验证集样本正确率最大值为89%,采用加权投票方法对分类器进行融合,得到各个分类器的权值,融合后的模型对蜂蜜的判别正确率提高到96%。该方法鉴别准速度快,确度高,适用于对蜂蜜品种的鉴别。
Abstract:A method for identifying honey varieties by near infrared spectroscopy based on multi-classifier fusion
was established. Using Fisher, SVM, PLS-DA and AdaBoost as single classifiers, the discriminant models of honey
varieties were established, and the single classifiers were screened by difference metrics to obtain the three most different
classifiers. The three single classifiers were fused and the model of multiple classifiers was used to distinguish and analyze
the honey varieties. The maximum accuracy of the honey validation set sample was 89% in a single classifier model. The
weighted voting method was used to fuse the classifier to obtain the weight of each classifier. The accuracy of the honey
was increased to 96 % in the combined model. The method is fast and accurate, it is suitable for the identification of honey
varieties.
关键词:蜂蜜;品种;多分类器;近红外光谱
Keywords:honey; varieties; multiple classifier; near infrared spectroscopy
基金:陕西省体育局项目(2018061)
本文引用格式:
张林,韩美林,杨琳,等. 基于多分类器融合的近红外光谱技术鉴别蜂蜜品种[J]. 化学分析计量,2019,28(3):38-41.
ZHANG L,HAN M L,YANG L,et al. Identification of honey varieties by near infrared spectroscopy[J]. Chemical analysis and meterage,2019,28(3):38-41.
相似文章: