Automated Tomato Quality Assessment Using Transfer Learning and Machine Learning Classifiers

Abstract

With the increasing demand for high-quality tomatoes and the need for efficient large-scale production, an automated grading system has become essential. Manual sorting is time-consuming, labor-intensive, and expensive, making automation a practical alternative. This study introduces a hybrid strategy that blends machine learning with deep learning methods for tomato classification. A custom dataset was created using specialized imaging hardware, followed by preprocessing techniques to enhance feature recognition. While classifiers including Support Vector Machines (SVM), Random Forest (RF), and K-Nearest Neighbors (KNN) were used for grading, For feature extraction, a Convolutional Neural Network (CNN) was employed. High classification accuracy was established by the suggested CNN SVM model, effectively distinguishing between healthy and defective tomatoes, as well as categorizing them based on ripeness levels. When tested against benchmark datasets, It functioned more accurately and efficiently than other hybrid models. Key performance parameters, such as accuracy, precision, recall, specificity, and F1-score, were used to further assess the model's efficacy.