Colour Classification of Harumanis Mango Quality using Artificial Neural Network

Abstract

Harumanis mango, a premium variety originating from Perlis, Malaysia, are only produced during limited seasons, making reliable and efficient quality evaluation highly important. Conventional inspection methods are typically performed manually, which can be time-consuming and prone to subjective judgment. This study proposes an automated system for classifying Harumanis mango quality by using Artificial Neural Network (ANN). High-resolution mango images were acquired using a CCD camera, and important features, particularly colour maturity, were extracted through HSI-based hue analysis. Initial classification was conducted using K-Nearest Neighbours (K-NN) and Support Vector Machine (SVM), achieving accuracy levels of up to 97.6% for colour-based maturity assessment. To further improve classification performance, a low-level data fusion strategy was applied, which combines multiple colour-related features. This fusion-based model produced a higher accuracy of 98.6%, surpassing the performance of individual classifiers. The findings indicate that combining visual features with ANN techniques enhances the accuracy and consistency of Harumanis mango quality classification. The proposed method provides a reliable, non-destructive and scalable solution for automated grading, minimizing dependence on manual inspection while improving overall quality control processes.