| In view of the important edible and medicinal value of Zanthoxylum,its quality control and grade division have gradually become an important part of the Zanthoxylum industry.Nowadays,the quality detection and classification criteria of Zanthoxylum are basically determined by the appearance,and the method is basically based on artificial sensory evaluation technology.However,artificial sensory evaluation has the inevitable defects such as subjective influence,evaluation fatigue and training difficulty,which makes it difficult to form a unified standardization of quality control and grade division of Zanthoxylum,and restricts the development of Zanthoxylum industry.Therefore,how to quickly,stably and standardly detect the quality of Zanthoxylum has become an urgent problem in the Zanthoxylum industry.In recent years,with the rapid development of intelligent sensory detection equipment and technology,it has become possible to detect the quality of Zanthoxylum intelligently.Therefore,this paper intends to target four groups of researches: red Zanthoxylum from six kinds of producing area,Sichuan blue/red Zanthoxylum,eight kinds of commercially available Zanthoxylum oil and seven different Zanthoxylum content of Zanthoxylum oil.As the electronic nose and electronic tongue for intelligent sensory detection equipment.,two kinds of eigenvalues are obtained based on eigenvalue optimization and Filter-Wrapper eigenvalue screening.With 22 model analysis methods such as support vector machine(SVM),K-nearest neighbor(KNN),decision tree and ensemble algorithm,the research focused on the topic of ‘Intelligent detection of the quality of Zanthoxylum and Zanthoxylum oil'.The main research and conclusions are as follows:(1)The electronic nose detection method for Zanthoxylum and Zanthoxylum oil was established.As object of research,four groups of Zanthoxylum and Zanthoxylum oil were detected,after using electronic nose sensor optimization and Filter-Wrapper eigenvalue screening and obtaining two types of eigenvalues,it was used to verify and analyze with 22 models.Compared the best accuracy of the electronic nose sensor optimization with the eigenvalue screening verification model: for the six kinds of producing area of red Zanthoxylum,the whole classification is 40.0%,Sichuan red Zanthoxylum is 83.3%,Shandong red Zanthoxylum is from 88.9% to 96.7% Shanxi red Zanthoxylum is 83.3%,Hebei red Zanthoxylum changed from 85.6% to 84.4%,Gansu red Zanthoxylum is from 97.8% to 92.2%,Shaanxi red Zanthoxylum oil is from 81.1% to 85.6%;for Sichuan green red Zanthoxylum,Subspace Discriminant and Linear SVM is 86.7%,Subspace KNN changed from 83.3% to 86.7%;for eight kinds of commercially available Zanthoxylum oil,the whole classification changed from 74.2% to 75.0%,Hanyuan Zanthoxylum oil is 96.7%,Dahongpao red Zanthoxylum oil were both 100.0%,Shushangxian Zanthoxylum oil changed from 100.0% to 99.2%,the Jinyangteng Zanthoxylum oil changed from 92.5% to 90.8%,the fresh rattan Zanthoxylum oil changed from 99.2% to 98.3%,and the special fresh Zanthoxylum oil is from 91.7.% to 92.5%,fragrant Zanthoxylum oil changed from 91.7% to 90.0%,and fresh rattan Zanthoxylum oil changed from 90.8% to 89.2%;for seven kinds of Zanthoxylum content,the accuracy rate was about 86%.(2)The electronic tongue detection method for Zanthoxylum and Zanthoxylum oil was established.As object of research,four groups of Zanthoxylum and Zanthoxylum oil were detected,after using electronic tongue sensor optimization and Filter-Wrapper eigenvalue screening and obtaining two types of eigenvalues,it was used to verify and analyze with 22 models.Compared the best accuracy of the electronic tongue sensor optimization with the eigenvalue screening verification model: for the red Zanthoxylum from six kinds of producing area,the whole classification is from 83.3% to 90.0%,Sichuan red Zanthoxylum is from 93.3% to 94.4%,Shandong red Zanthoxylum is from 91.1% to 97.8% Shanxi red Zanthoxylum is 100.0%,Hebei red Zanthoxylum changed from 91.1% to 100.0%,Gansu red Zanthoxylum is from 91.1% to 96.7%,Shaanxi red Zanthoxylum oil is 100.0%;for Sichuan green red Zanthoxylum,the accuracy of Fine Tree is 96.7%,Linear SVW and Cubic KMN is 100%;for eight kinds of commercially available Zanthoxylum oil,the whole classification changed from 82.5% to 81.7%,Hanyuan Zanthoxylum oil is 100.0%,Dahongpao red Zanthoxylum oil is from 96.7% to 95.0%,Shushanxian Zanthoxylum oil is 100.0%,the Jinyangteng Zanthoxylum oil changed from 100.0% to 99.2%,the fresh rattan Zanthoxylum oil changed from 92.5% to 91.7% and the special fresh Zanthoxylum oil is from 95.0 % to 93.3%,fragrant Zanthoxylum oil changed from 98.3% to 95.8%,and fresh rattan Zanthoxylum oil changed from 99.2% to 96.7%;for seven kinds of Zanthoxylum content,the accuracy rate was about 86%.(3)Merged electronic nose and tongue sensor optimization eigenvalue,then,after using Filter-Wrapper eigenvalue screening and obtaining two types of eigenvalues,it was used to verify and analyze with 22 models.Compared the best accuracy of sensor optimization with eigenvalue screening verification model:(4)The comprehensive comparison shows that the best results were obtained by using the electronic tongue detection and SVM model for the red Zanthoxylums of the six kinds of producing area;this three models,the electronic tongue detection and decision tree,KNN and SVM were the best for Sichuan green red Zanthoxylum;As for the eight kinds of commercially available Zanthoxylum oil,using electronic nose-electronic tongue detection and SVM and ensemble algorithm model is the best;As for seven kinds of Zanthoxylum content Zanthoxylum oil,the consequence of using electronic nose,electronic tongue,the combined detection and 22 models analysis were very similar. |
PDF Full Text Request