Rapid And Nondestructive Detection Of Fusarium Graminearum And Its Toxins In Wheat And Maize Based On Multispectral Imaging Technology

Cereal is a necessary material for human survival,but it is easy to be polluted by mycotoxins and causes serious harm to human society.Therefore,the detection of mycotoxins in cereals has become an important direction of scientific research and practitioners.Traditional detection methods of cereal mycotoxins are complex and costly,which can not achieve rapid and non-destructive detection of cereal mycotoxins.As a new detection technology,multispectral imaging(MSI)has many advantages,such as simple,fast,non-destructive,and can obtain the image and spectral information of the measured substance at the same time.Based on MSI,the rapid and non-destructive detection of Fusarium graminearum and its toxins in wheat and maize was studied.(1)Rapid detection of Fusarium graminearum content and growth status in wheat based on multispectral imaging technology.Through artificial infection of wheat,the samples cultured for 0,3,6,9,15,18 and 21 days after infection were taken for multispectral imaging and fungal content detection,and different modeling methods were used to complete the prediction of Fusarium graminearum content in wheat.The results show that the correlation coefficient(R)and root mean square error(RMSE)of genetic algorithm-support vector machine(GA-SVM)in calibration set and prediction set were 0.9663,0.9292,0.5992 CFU/g and 0.6725 CFU/g,respectively;The accuracy of genetic algorithm-back propagation neural network(GA-BPNN)was as high as100 %.(2)The rapid nondestructive detection of ZEN in maize based on multispectral imaging technology.The prediction results of ZEN content in natural infected maize showed that the R of calibration set and prediction set using GA-BPNN were 0.9567 and 0.9507,and the RMSE were 0.6866 and 0.6613 ?g/kg,respectively.The results show that the best GA-BPNN model can classify maize samples with different pollution degree,and the classification accuracy was as high as 93.33 %,which indicates that this method can detect ZEN content in maize well.(3)The content and pollution level of DON in wheat were identified based on multispectral imaging.In this study,the naturally infected wheat was taken as the research object.In the prediction of DON content in wheat,the prediction results of calibration set and prediction set by GA-SVM method R and RMSE were 0.9923,0.9988,215.1 ?g/kg and 365.3 ?g/kg,respectively.In the identification of DON pollution level in wheat,the accuracy of prediction set(A %)of principal component analysis-partial least square(PCA-PLS)modeling was 94.29 %,which realized the identification of different DON pollution levels in wheat.At the same time,the differences of DON content in different parts of wheat(upward and downward)were further studied.When the pollution level was low,the reflection intensity of downward was greater than the reflection intensity of upward.However,with the increase of pollution degree,the reflection intensity of upward gradually approached,and finally was greater than that of downward.This research shows that the rapid and non-destructive detection of cereal fungi and its toxins can be realized based on MSI,which provides a new method for the monitoring of cereal mycotoxins pollution in the process of production,processing and storage.