Automated Tomato Classification using CNN-ELM with Enhanced Whale Optimization for Grading

Tomato grading is essential in lowering production costs without compromising product quality in agricultural practices. However, manual sorting is time-consuming and mistaken. This study proposes an automated system for tomato classification into three categories: Ripe, Unripe, and Reject, using a Convolutional Neural Network (CNN) integrated with an Extreme Learning Machine (ELM) classifier, optimized by an Enhanced Whale Optimization Algorithm (EWOA). The CNN extracts relevant features from the images, while the ELM efficiently classifies the tomatoes based on these features. The EWOA optimizes key hyperparameters to improve the efficiency of the model. The data set utilized for analysis and training consisted of 800 images per class, split into 80% for training and 20% for analysis. The model achieved an overall accuracy of around 80%, with a particularly strong performance in classifying Reject tomatoes, where 150 out of 155 were correctly identified. However, the model exhibited some confusion matrix between the Ripe and Unripe classes, as 57 unripe tomatoes were misclassified as ripe. The training and validation loss and accuracy metrics showed consistent improvement over 100 epochs, demonstrating the model’s ability to generalize effectively.