Detection of fungal infection in pistachio kernel by long-wave near-infrared hyperspectral imaging technique

Hyperspectral imaging is a non-destructive technique with great capability to detect food defects and diseases. In this research, infection of pistachio kernels by two different isolates of Aspergillus flavus, KK11 and R5, was determined using long-wave near-infrared hyperspectral imaging. Seven fungal growing stages on pistachio kernels, for both isolates, were investigated. The features from the hypercubes of healthy and infected pistachios were extracted, selected and then used for classification by linear discriminant analysis and quadratic discriminant analysis (QDA) methods which were performed by 10-fold cross validation technique in MATLAB 2010a. The QDA model gave the highest classification accuracy for all classes (healthy and infected kernels by different fungi and at different stages). For classification of infection by different fungi at the last fungal growing stage, QDA had the accuracy of 91.7%. As KK11 is an aflatoxin-producing and R5 is a non-aflatoxin-producing fungus isolate, the importance of the technique to detect aflatoxin contamination in pistachio is significant.