摘要
利用高光谱成像技术进行玉米籽粒表面黄曲霉毒素的检测.将黄曲霉毒素原液用甲醇分别稀释成浓度为10、20、100、500 μg/L的溶液,然后逐一滴在等量4组共120粒玉米籽粒的表面,取同品种30粒洁净玉米籽粒作为对照组.利用400～1 000 nm可见/近红外高光谱成像系统进行样品图像光谱信息获取,采用标准正态变量校正进行数据预处理.首先经主成分分析（Principal Component Analysis,PCA）进行高光谱数据降维,然后利用PCA从835个波长中提取的前14个主成分为输入,采用因子判别分析（Factorial DiscriminantAnalysis,FDA）对5类样品进行分类.FDA构建的模型对训练集和验证集的判别准确率分别达95％和86％.结果表明利用高光谱成像技术并结合PCA-FDA方法进行玉米籽粒表面黄曲霉毒素的检测是可行的.
The detection of aflatoxin on the surface of corn granule by using hyperspectral imaging technology was discussed. The aflatoxin stoste was diluted with methanol to four different concentrations of aflatoxin solutions, 10, 20, 100 and 500 p,g/L. These solutions were dripped on the corn granule surface of four groups of which every group consisted of 30 intact kernels. Also the other 30 clean kernels were selected as control samples. A VIS/NIR hyperspectral imaging system with spectral range of 400 - 1 000 nm was established to acquire spectral data, and then these data were preprocessed by standard normal variate correction. Firstly, principal component analysis was execu- ted to reduce the dimensionality. Then the first 14 principal components which were selected from the original 835 wavelengths by PCA were used to establish the FDA discrimination function to classify the 5 kinds of samples. Classi- fication accuracies of 95% and 86% for the training and testing sets were achieved by using FDA model. The result indicated that it was feasible to identify and classify aflatoxin on the corn granule surface by using the hyperspectral imaging technology and method proposed in this paper.
出处
《中国粮油学报》
EI
CAS
CSCD
北大核心
2014年第12期107-110,118,共5页
Journal of the Chinese Cereals and Oils Association
基金
国家科技支撑计划(2012BAK08B04)
关键词
玉米颗粒
黄曲霉毒素
因子判别分析
主成分分析
高光谱成像
corn granule, aflatoxin, factor discriminant analysis, principal component analysis, hyperspectralimaging
