Study On Detection Methods Of Corn Quality And Safety Based On Olfactory Visualization Technology

Corn is an important crop in China,when the quality and safety of corn are jeopardized,it can affect people’s health.In this study,moisture content was used as the quality index of corn and Aflatoxin B₁（AFB₁）concentration was used as the safety index.The existing techniques for detection of moisture content and AFB₁concentration in corn have disadvantages such as complicated operation,high cost and long detection time.Therefore,this study investigated the quantitative detection method of moisture content and AFB₁ based on olfactory visualization technology with corn as the research object.The specific research contents are as follows:（1）Study on corn moisture content detection method based on olfactory visualization technology.A home-made olfactory visualization detection system was used to obtain volatile gas information of corn samples and characterize them in the form of images;the RRelief F algorithm was used to rank the feature weights of the color components,combined with support vector regression（SVR）to determine the best color component combinations;while comparing the grid search（GS）and particle swarm optimization（PSO）algorithms optimization（PSO）to optimize the key parameters of the model;finally,a quantitative detection model was developed to achieve rapid detection of moisture content of stored corn.The results of the model operation show that compared with the GS-SVR model,the correlation coefficient of prediction（R_P）of the RRelief F-GS-SVR model is improved from 0.97 to 0.98,and the root mean square error of prediction（RMSEP）decreased from 0.66%to 0.53%;compared with the PSO-SVR model,the R_P of the RRelief F-PSO-SVR model improved from 0.98 to 0.99,and the RMSEP decreased from 0.59%to 0.48%.The overall results show that the use of olfactory visualization technology can achieve rapid quantitative detection of corn moisture content;in addition,feature screening and model parameter optimization can further improve the detection accuracy and generalization performance of the model.（2）Study on corn AFB₁ content detection method based on olfactory visualization technology.A feature selection approach based on multi-subspace randomization and collaboration（SRCFS）algorithm combined with support vector machine regression and back propagation neural network（BPNN）model to optimize the color component features to determine the best color component combination;the SVR and BPNN models based on the optimized color components were developed to achieve rapid quantitative detection of corn AFB₁,during the calibration of the SVR model,the particle swarm optimization algorithm was used to optimize the parameters of the SVR model.The results of model operations showed that the SVR model outperformed the BPNN model;after feature screening,the mean R_P value of the PSO-SVR model increased from 0.9654 to 0.9837 and the mean RMSEP value decreased from 4.8045μg·kg^-1 to 3.5222μg·kg^-1;the R_P and RMSEP of the best SRCFS-PSO-SVR model were 0.9898 and 2.8069μg·kg^-1,respectively.The results indicated that the use of olfactory visualization technique could achieve the quantitative detection of AFB₁ concentration in corn with high accuracy.This study demonstrates the application value of olfactory visualization technology in the detection of corn moisture content and AFB₁ concentration.The research results can provide a technical reference and theoretical basis for the design and development of olfactory visualization detection devices for grain quality and safety detection,and also provide a new,fast and accurate detection method for grain bin managers.