Classification of strawberry ripeness stages using machine learning algorithms and colour spaces

Abstract

Accurate classification of strawberry ripeness is a crucial aspect of ensuring high-quality food products, optimizing harvesting and storage processes, and promoting consumer health. Although several non-destructive computer vision-based systems have been developed for this purpose, the influence of different colour spaces on machine-learning model performance during the ripeness stage classification of strawberries remains underexplored. In this context, three machine-learning models, namely Gaussian Naive Bayes (GNB), support vector machine (SVM) and feed-forward artificial neural networks (FANN), were combined with four colour spaces (RGB, HLS, CIELab and YCbCr) and biometrical characteristics to evaluate the effectiveness of colour spaces on the performance of machine-learning models for classifying strawberry ripeness. For this purpose, 1210 samples were collected and manually classified into four ripeness stages. A dataset was created by combining each colour space value, biometrical properties, and corresponding ripeness stage, which was used as inputs to the models. The results indicated that FANN with CIELab colour space achieved the highest accuracy of 96.7%, followed by GNB and SVM, both having equal accuracy of 95.46% in CIELab colour space. The least accuracy of 92.15% was observed in RGB colour space with the GNB classifier. In this study, the unripe and over-ripe stages were more accurately classified, while intermediate ripening stages proved to be more challenging for the models. Furthermore, the accuracy of models was observed to be influenced by both the colour space and classification model selected. Additionally, further research is needed to investigate other features that could improve the performance of models for strawberry ripeness classification.