Development of machine vision based system for classification of Guava fruits on the basis of CIE1931 chromaticity coordinates

Abstract

The present work represents a non contact, machine vision based method to estimate ripeness level of Guava fruit. The fruit under test is classified as green, ripe, overripe and spoiled using a web camera based computer vision system. This simple method uses a combination of digital web camera; computer and indigenously developed GUI based software to measure and analyze the surface color of the fruits. The images of the fruit under test are grabbed and displayed on computer screen. Quantitative information such as RGB color distribution, CIE1931 standard based tristimulus values, Chromaticity coordinates and averages (in terms of L_, a_ and b_ values) are computed. The developed software appropriately analyzes the color of the fruit skin and is also capable of classifying the ripening stage of the guava fruit as green, ripe, overripe and spoiled using Principal Component Analysis (PCA). Distinct clusters for the ripeness classes were readily observed in the PCA scatter plot. An Artificial Neural Network (ANN) was also used for a better prediction for unknown samples.