Processing the Hyperspectral Images for Detecting Infection of Pistachio Kernel by R5 and KK11 Isolates of Aspergillus flavus Fungus

Hyperspectral imaging technique as a new and efficient method is applied for detecting infection in agricultural products. It was used for classification of healthy and infected pistachio kernels by Aspergillus flavus fungus with and without considering infection stages. Two different fungus isolates, R5 and KK11 with and without capable of producing aflatoxin, respectively, were individually used to infect the pistachio kernel samples. From 960 to 1700 nm, three effective wavelengths of 1090, 1280, and 1700 nm were selected by principle component analysis method. After feature extraction, K-fold cross validation, support vector machine, and artificial neural network methods were used for classification. The results showed that the classification accuracy of the K-fold cross validation method was higher for classifying the healthy and infected pistachios without considering the infection stages and isolate type (99.71%). The maximum accuracy of the developed algorithms in classification of isolate type and infection stage was obtained as 69-91%.