高光谱亚像元分解预测花生中的黄曲霉毒素B1

Detecting Aflatoxin B1 in Peanuts by Hyperspectral Subpixel Decomposition

黄曲霉毒素B1是一种剧毒、强致癌物质,具有紫外荧光特性。为研究高光谱成像技术对黄曲霉素的检测能力,在365 nm紫外灯下,通过高光谱成像系统采集5个浓度共250个花生籽粒样本33个波段(400~720 nm)的高光谱图像。提出一种基于高光谱亚像元分解丰度图像直方图量化特征预测黄曲霉毒素含量的方法。该方法首先通过N-FINDR端元提取方法获得黄曲霉毒素端元光谱,然后对高光谱图像进行非负矩阵分解(NMF),得到黄曲霉毒素丰度图像,对丰度图像构建直方图量化特征,使用偏最小二乘回归(PLS)和支持向量机回归(SVR)进行黄曲霉毒素丰度反演,五折交叉验法得到平均两种回归模型预测相对误差分别为29.95%和12.16%,RMSE最高为0.0306。本研究结果对农产品籽粒黄曲霉毒素光学快速检测具有积极意义。

Aflatoxin is a highly toxic and carcinogenic substance with UV fluorescence characteristics.To study the detection of aflatoxins by hyperspectral imaging,aflatoxin was collected by a hyperspectral imaging system at 365 nm UV light.A total of 250 peanut grain samples in the 33 bands(400-720 nm)hyperspectral images.A method for predicting aflatoxin content based on the histogram quantization features of hyperspectral image decomposition abundance images was proposed.The method first obtained aflatoxin end-band spectra by N-FINDR endmember extraction,Spectral images were subjected to non-negative matrix factorization(NMF)to get aflatoxin abundance images.Based on this image,histogram quantization features were constructed.Partial least squares regression(PLS)and support vector machine regression(SVR)Vegetation abundance inversion,50%cross-validation method obtained the average relative error of the two regression models respectively 29.95%and 12.16%,RMSE up to 0.0306.The results of this study have positive significance for the optical rapid detection of aflatoxin in agricultural products.