AUTOMATED DETECTION OF PISTACHIO DEFECTS BY MACHINE VISION

Abstract

Pistachio (Pistacia vera) nuts with shell and kernel defects detract from consumer acceptance and, in some cases, may be more prone to insect damage, mold decay, and/or aflatoxin contamination. The objective of this study was to develop imaging algorithms to improve sorting of nuts with the following shell defects: oily stains, dark stains, adhering hull, and the following kernel defects: navel orangeworm (NOW) damage, fungal decay, and Aspergillus molds, all of which indicate risk of aflatoxin contamination. Imaging algorithms were developed to distinguish normal nuts from those nuts with oily stains, dark stains, and adhering hull as well as nuts having kernel defects. Image algorithm testing on a validation data set showed that nuts having oily stain, dark stain, or adhering hull could be distinguished from normal nuts with an accuracy of 98%. Removing nuts with oily stain, dark stain, and adhering hull will also remove 89.7% of nuts with kernel decay, 93.8% of nuts with Aspergillus molds present, and 98.7% of NOW positive nuts.