A vision system to identify aflatoxin contaminated cashews, pistachios and brazil nuts

Fluorescence techniques have shown great potential for detecting aflatoxin contaminated foods, since this metabolite can excrete a characteristic substance (Kojic acid) which is fluorescent after ultra violet illumination. In this paper we present the results of the application of a fluorescent sorting method (6 Wood lamps, 18 W each,  = 365 nm) based on an artificial vision system applied to cashews, pistachios and brazil nuts. The final purpose is to detect defects related to changes on the nuts surface standard conditions due to the aflatoxin contamination. Therefore 146 monochromatic images of the samples (73 contaminated and 73 uncontaminated) were acquired in the optical range between 410 and 600 nm with a vision system endowed with different selective filters; for each image the gray level of the region of interest (ROI) was evaluated and then a statistical analysis was developed with the purpose to select the optimal optical range to separate contaminated and uncontaminated nuts. Afterwards a multispectral analysis, conducted with the creation of level curves, was set up to identify the most useful combination of wavelengths to classify the samples (whole pistachios: 560-600 nm, pistachios kernel without shell: 520-410 nm, cashews: 520-410 nm, brazil nuts: 560-600 nm).