Machine vision grading of pistachio nuts using gray-level histogram

A machine vision system was used to classify "in the shell" pistachio nuts based on USDA grades. The gray-level histogram data obtained from the gray scale image of the nuts were analyzed to select a set of suitable recognition features. Based on the analyses, the mean of the gray-level histogram over 50 to 60 gray-level range and the area of each nut (the integral of its gray-level histogram) were selected as the recognition features. The selected features were used as input to three classification schemes: a Gaussian, a decision tree, and a multi-layer neural network (MLNN). The three classifiers had similar recognition rates. However, the MLNN classifier resulted in slightly higher performance with more uniform classification accuracy than the other two classifiers.