Date inspection by color machine vision

Abstract

A color computer vision system consisting of a microcomputer with an image frame grabber and a CCD color camera was developed for sorting and grading Saudian dates (Sifri variety) based on a color thresholding technique. Samples of dates graded by experts were used for the experiment. Histograms of the RGB color bands were tested. Histogram features mean, variance, and wrinkle ratio were used as the criteria for classification of dates into four quality classes. The system showed an average classification error of 1.8% using features from the red color band.