Research On Artificial Electronic Tongue Based On Ensemble Learning And Its Application In Agricultural Product Quality Detection

With the improvement of people's living standards,people have higher requirements for the quality of agricultural products.Therefore,An efficient,convenient and objective method for agricultural product quality detection is urgently needed which is efficient,convenient,accurate and objective.As a new type of detection technology,the electronic tongue intelligent sensory detection system can obtain the "fingerprint information" from the liquid sample by combining pattern recognition methods.However,it is difficult to improve the predictive performance of traditional pattern recognition methods.Therefore,this dissertation developed a voltammetric electronic tongue system and studied its application in the quality of agricultural products based on virtual instrument technology and ensemble learning methods.The specific research contents are as follows:(1)A set of intelligent sensory electronic tongue system is designed which is based on virtual instrument technology.It has the advantages of small size,low cost,convenient detection and no pre-processing.The system consists of a sensor array,a signal conditioning module,a data acquisition card,and a host computer software.(2)Aiming at the shortcomings of single classifiers commonly used in electronic tongue pattern recognition with low generalization performance and low classification accuracy,this dissertation proposes to use the idea of ensemble learning.It fuse the model of single learner effectively and construct different ensemble learning models for discriminant analysis of lemon slices with four different processing technologies.Discrete wavelet transform is used to preprocess the electronic tongue signal,and then based on BP neural network,K-nearest neighbor algorithm and logistic regression algorithm as the base classifier,the Bagging,Boosting and Stacking ensemble learning models are separately constructed to characterize lemon slices with different processing techniques.The experimental results show that the classification effect of the ensemble learning method is superior to the traditional machine learning method.(3)Aiming at the problem that the traditional ensemble learning algorithm does not consider the different sample data and the adaptability of the base classifier,the idea of dynamic weight is introduced,and the CELEDAT algorithm is proposed for qualitative analysis of aged wheat with different storage years.The wavelet packet transform was used to preprocess the data,and then cluster algorithms such as genetic algorithm,particle swarm algorithm and artificial fish swarm algorithm are used to optimize the parameters ofsingle classifiers such as BPNN and LSSVM.The CELEDAT model was constructed based on the parameter-optimized BPNN,LSSVM,RF,and ELM as the base classifier.The final experimental results show that compared with the ensemble learning models of Bagging,Boosting,and Stacking,the CELEDAT model has better classification and recognition effect on aging wheat with different storage years.