Research On Confirmation Model Of Rice Field Of Liuhe Based On Artificial Neural Network

As the distinctive geographical location and climate in Jilin Province,it has been known throughout the country for its geographically-recognized rice with multiple advantages and unique qualities.It is of great significance to study the technology for confirming the geographical indication of rice in Jilin Province.This paper mainly discusses the feasibility of applying Back-Propagation Artificial Neural Network(BP-ANN)to confirm the origin of the adjacent areas,and establishes the confirmation model for the non-Liuhe rice origin in Liuhe and adjacent areas.It will provide a theoretical basis for the establishment of the Jilin Province geographical indication rice protection system.In this study,120 samples of rice were collected from Liuhe,Jilin Province and its adjacent non-Liuhe area.The varieties were all Daohuaxiang.Atomic absorption spectroscopy was used to detect Cu,Zn,Fe,Mn,K,Ca,Na,Mg,Pb,Cd and 10 mineral elements in rice samples.At the same time,Fourier spectrum near-infrared spectroscopy was used to collect spectral data.The obtained data were preprocessed separately,and the training set and test set were divided.The BP neural network model was established respectively,and the established model was optimized by using genetic algorithm(GA).The main conclusions are as follows:(1)The BP-ANN origin model established based on mineral element data can accurately classify the Liuhe and non-Liuhe producing areas with high accuracy of identification,and the classification accuracy of the training set and the test set was 95% and 85% respectively.(2)The BP-ANN production origin confirmation model based on mineral elements was optimized by weight threshold of GA,which improved the precision of the model.The accuracy of the optimized training set was 96.25%,which was 1.25% higher than that before optimization.The accuracy of the test set was 87.5%,which was 2.5% higher than that before optimization,and the generalization ability is also better.(3)Near-infrared spectroscopy data were preprocessed by Savitzky-Golay convolution smoothing.Principal component analysis was used to determine which data to reduce dimension.The first three principal components were extracted as input variables,and the BP-ANN model was established.The accuracy of model training set was 97.83% and that of test set was 87.5%.The confirmation of the origin of the Liuhe geographical indication rice can be accurately performed.(4)Based on near-infrared spectroscopy,the BP-ANN corroboration model of origin has been optimized by GA.The accuracy of the model training set can reach 100%,the test set accuracy can be increased to 91.3%,the model specificity can reach 100%,and the generalization ability is better than before the optimization.(5)The combination of GA and BP-ANN model is feasible in the confirmation of origin,and GA can make the generalization ability of the model better,which the accuracy of origin identification being higher.