Skip to main content
ARS Home » Southeast Area » Tifton, Georgia » Crop Genetics and Breeding Research » Research » Publications at this Location » Publication #339219

Research Project: Genetic Improvement of Maize and Sorghum for Resistance to Biotic Stress

Location: Crop Genetics and Breeding Research

Title: Detection of aflatoxin B1 (AFB1) in individual maize kernels using short wave infrared (SWIR) hyperspectral imaging

Author
item Chu, Xuan - China Agriculture University
item Wang, Wei - China Agriculture University
item Yoon, Seung-chul
item Ni, Xinzhi
item Heitschmidt, Gerald - Jerry

Submitted to: Biosystems Engineering
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/12/2017
Publication Date: 3/6/2017
Citation: Chu, X., Wang, W., Yoon, S.C., Ni, X., Heitschmidt, G.W. 2017. Detection of aflatoxin B1 (AFB1) in individual maize kernels using short wave infrared (SWIR) hyperspectral imaging. Biosystems Engineering. 157:13-23.

Interpretive Summary: As one of the systematic research efforts on utilizing hyperspectral imaging technique to detect aflatoxin contaminations on maize kernels, the short wave infrared hyperspectral imaging technique (with wavelength of 1000-2500 nm) was utilized to detect aflatoxin B1 in the individual maize kernels. Thirty maize kernels from each of the four varieties that had been artificially inoculated with a toxigenic strain of Aspergillus flavus were harvested from the field and then examined. Multivariable data analyses showed that the shortwave hyperspectral imaging can qualitatively classify the aflatoxin contamination levels (<20 ppb, 20-100 ppb, 100 ppb) in all individual kernels without effect of four maize varieties. Classification accuracies were 83.75% and 82.50% for calibration and validation sets, respectively. In addition, a general positive correlation exists between categorical aflatoxin content and the first three principal components. Coefficients of determination of the support vector machine regression model were 0.77 and 0.70 for calibration and validation sets, respectively. In addition, five wavelengths (1317, 1459, 1865, 1934 and 2274 nm) were selected as characteristic wavelengths to detect aflatoxin contaminations on maize kernels. Results indicated that hyperspectral imaging could be used to classify aflatoxin level qualitatively in individual maize kernels, but the precision of predicting the categorical aflatoxin content still needs to be further improved.

Technical Abstract: Short wave infrared hyperspectral imaging (SWIR) (1000-2500 nm) was used to detect aflatoxin B1 (AFB1) in individual maize kernels. A total of 120 kernels of four varieties (or 30 kernels per variety) that had been artificially inoculated with a toxigenic strain of Aspergillus flavus and harvested from the field were examined using the SWIR technology. Normalization and principal component analysis (PCA) were applied on average spectra of each kernel to reduce dimensionality and noise. Combining with support vector machine classification methods, the first five principal components (PCs) were used to qualitatively classify the AFB1 contamination levels (<20 ppb, 20-100 ppb, 100 ppb) in single kernels without effect of maize variety. Classification accuracies were 83.75% and 82.50% for calibration and validation sets, respectively. It was also noted that a general correlation exists between categorical AFB1 content and the first three PCs. Coefficients of determination (R2) of the support vector machine regression model were 0.77 and 0.70 for calibration and validation sets, respectively. A possible distribution map of AFB1 was also made by applying the regression model on every pixel of the hyperspectral image. Moreover, using loading plots of the mutual first three PCs, five wavelengths (1317, 1459, 1865, 1934 and 2274 nm) were selected as characteristic wavelengths. Results indicated that hyperspectral imaging could be used to classify AFB1 level qualitatively in individual maize kernels. However, the performance of predicting the categorical AFB1 content still needs to be improved.