Study On Origin Identification Of Star Anise Based On Electronic Nose Technology

The origin identification and grade classification of anise are usually realized through sensory analysis.However,sensory analysis has some shortcomings such as strong subjectivity,poor repeatability,low accuracy,and lack of scientific and normative identification results.Electronic nose is an odor detection system designed to simulate human’s odor identification mechanism,which has the advantages of convenience and high accuracy.Therefore,this paper uses electronic nose to identify the origin of star anise.(1)The hardware system of electronic nose for the origin identification of star anise was designed.The hardware system mainly includes Ardunio,air pump,solenoid valve,LCD display module,sensor array and air chamber.Using ANSYS to carry on the simulation analysis of two kinds of gas chamber,according to the advantages and disadvantages of different structure of the air chamber to optimize the suitable air chamber structure.Complete the whole machine assembly and collect data.There are a total of 150 samples of star anise samples from 3 producing areas.Each group of samples contains the output values of 14 gas-sensitive sensors.Data from 14 sensors are read for 50 times during each sample collection as the recorded values of the test samples.After data collection,the data set matrix is 150×14×50.(2)Gas sensor array selection.First,Fourteen gas sensors were selected according to the odour components of anise,and then the highly correlated sensor groups were found and selected respectively by combining the coefficient of variation and single-factor variance to obtain two sensor arrays.Finally,the decision tree algorithm was used to compare and analyze the results which showed that the 12 gas sensor arrays selected by the coefficient of variation had the best identification effect,and the 12 gas sensor arrays selected constituted the original data set matrix150×12×50.(3)Feature selection and dimension reduction.Firstly,the mean,maximum and minimum values of 50 sensor data were extracted to obtain three new feature matrices(150×12),and then the LDA algorithm was used to compare and analyze the three feature matrices respectively.The results show that the distance between the projection points of the eigenmatrix samples constituted by the mean value is the largest,and the distance within the class is the smallest,that is,the effect is the best.Finally,PCA algorithm is used to reduce the dimension of the mean feature matrix,and the size of the obtained feature matrix is 150×9.(4)The classification model is established.Based on the feature extraction and dimensionality reduction data,ELM,GSA-SVM,GA-SVM and PSO-SVM methods are used to identify and analyze the origin of star anise.The results show that the PSO-SVM classification model has the highest average accuracy rate,which can reach 93.33%.which indicates that this model can better realize the origin identification of anise.