Designing of guava quality classification model based on ANOVA and machine learning.

The precise maturity quality classification of guava is crucial at farm level, retail, storage, and supply chain. The manual classification causes substantial postharvest losses, which increases the demand for material and resources. Therefore, the present study proposed an automated, precise, and accurate model for quality classification according to the maturity stages (Green, Mature Green, Ripe) of three varieties (Local Sindhi, Riyali, Thadhrami) of guava. The study aimed to develop a precise, accurate and automated model for the classification of guava according to their maturity stages. The guava images were used to extract color, shape, and texture features. Analysis of Variance (ANOVA) was used for the selection of important features. The six different machine learning (ML) classifiers; Artificial Neural Network (ANN), k-Nearest Neighbor (KNN), Support Vector Machine (SVM), Cubic SVM, Quadratic SVM, and Random Forest (RF) were used to find out the best classifier for maturity classification. Among the proposed classifiers, the RF classifier was found to be the best classifier for all three varieties of guava. The Quadratic SVM classifier showed the lowest classification accuracy. The study concluded that RF classifier was found to be a robust model for the maturity classification of guava.