Study Of Moldy Wheat Detection Based On Colorimetric Sensor Array-Visible/near Infrared Spectrum Technology

During the storage,transportation and processing of wheat,it is easy to cause the germination and mildew of wheat due to the microorganism and some other factors of external environment.In the process of mildew,large amounts of organic matters will decompose to volatile substance by the growth and metabolism of microorganism.This study got the images and spectral data of the colorimetric sensors which has acted with the VOCs of wheat samples with different mildew degree,through analyzing the images and spectral data of the sensors,the qualitative and quantitative analysis of VOCs of mildew wheat can be accomplished.The main research contents of this paper is as follows:（1）The research of volatile organic compounds changes in the process of wheat mildew.GC-MS has been applied to the determination of the VOCs of wheat samples with different mildew degree（fresh wheat,storage in specific environment for 3,7,11days mildew wheat）.53 kinds of VOCs were detected.Under the principal component analysis of the 53 kinds of VOCs,it can be found that the contribution of 1-octene-3-alcohol is highest in the load factors of the former three principal components.And the content of 1-octene-3-alcohol was growing significantly with the increase of mildew level of wheat.That means 1-octene-3-alcohol can be used as the characteristic volatile compound of the degree of mildew wheat.（2）The research of mildew wheat detection based on olfactory visualization technology.The experiment using visualization system to extract specific color sensitive materials which sensitive to 1-octene-3-alcohol and the VOCs of wheat with different mildew degree from 30 kinds of color sensitive materials.Then,the specific color sensitive materials were used to build up CSA to accomplish the detection of mildew wheat.The experiment results show that,the CSA which composed of NO₂BDP,BrBDP and HBDP three fluorine boron pyrrole compounds can complete the accurate identification of different degree of mildew wheat by olfactory visualization system.After PCA of the fresh wheat,storage in in specific environment for 3,7,11 days mildew wheat samples,KNN and LDA pattern recognition were employed to analyze.Compared the KNN model with LDA model,it can be found that when K is 2,and the principal component number is 9,the KNN model is the best model for identification,training set recognition rate reached 95.83%,prediction set recognition rate reached95.83%.（3）The research of mildew wheat qualitative detection based on visible/near-infrared combined with color sensor technology.This experiments using visible/near-infrared spectrometer to acquire the spectral information of CSA which after reaction with the VOCs of different degree of mildew wheat,and through the processing and pattern recognition analysis of the spectrum data to distinguish different degree of mildew wheat.The research results show that,after the SNV processing and Si-PLS variable selection of the original visible/near-infrared spectra of NO₂BDP,BrBDP and HBDP sensor which after reaction with the VOCs of different degree of mildew wheat.PCA model which build up with the selected data can well separate the fresh wheat,storage in in specific environment for 3,7,11 days mildew wheat samples.In order to optimize variables again,the GA algorithm is adopted to select the data for 10 times again.After PCA of each group of selected data,extracted the principal component factors as input for corresponding KNN and LDA analysis respectively.It can be found that when K is 2,and the principal component number is 9,the KNN model is the best model for identification,training set recognition rate reached 100%,prediction set recognition rate reached 96.25%.（4）The research of mildew wheat quantitative detection based on visible/near-infrared combined with color sensor technology.This experiment selected ash green aspergillus as the research object,mildew colonies as an index,through the visible/near-infrared spectrum of CSA to reflect the differences between the samples,and select the corresponding characteristic wavelengths to establish quantitative analysis model of the total mold colonies.The research results show that,after the SNV processing with the original visible/near-infrared spectra of NO₂BDP,BrBDP and HBDP sensor which after reaction with the VOCs of different degree of mildew wheat,the PLS and Si-PLS model of total colonies were set up respectively with the corresponding spectrum data.By contrast,it can be found that Si-PLS model of each sensor is superior to corresponding PLS model.To further improve the detection precision of the total number of colonies,after joint of the characteristics spectrum interval of the NO₂BDP,BrBDP and HBDP sensor which selected by Si-PLS,GA algorithm was used for variable selection for 10times of the joint of the characteristics spectrum interval again,then PLS models were respectively built up with the 10 groups of variables.The best Si-GA-PLS model can be achieved according to the comparison of the 10 models,and the optimum result from Si-GA-PLS mode was with RMSEP=0.709lgcfu,and Rp=0.9387.Research results show that the quantitative detection of VOCs can be realized through the visible/near-infrared combined with color sensor technology,and this research has realized the quality control off mildew wheat.