Classification Of Palm Oil Fruit Maturity Using CNN And Multiclass SVM In Ara Bungong Village

Abstract

The determination of the ripeness level of fresh fruit bunches (FFB) of oil palm is still largely performed manually, making it susceptible to subjectivity and classification errors that can affect harvest quality and consistency. This study aims to develop an automated classification system for oil palm fresh fruit bunch ripeness by combining a Convolutional Neural Network (CNN) as a feature extractor and a Multiclass Support Vector Machine (SVM) as the classification algorithm. The dataset consisted of 300 images of oil palm fresh fruit bunches categorized into three classes: unripe, ripe, and overripe. The research stages included image data collection, image preprocessing, feature extraction using CNN, classification using Multiclass SVM, and performance evaluation based on the confusion matrix, accuracy, precision, recall, and F1-score. The CNN was employed to automatically extract representative visual features from the input images, while the Multiclass SVM classified the extracted feature vectors into the corresponding ripeness categories. The experimental results showed that the proposed model achieved an accuracy of 90%, precision of 90%, recall of 90%, and an F1-score of 90% in classifying the ripeness levels of oil palm fresh fruit bunches. These findings indicate that the combination of CNN and Multiclass SVM effectively recognizes the visual characteristics of each ripeness level and provides reliable classification performance. The developed system is expected to serve as an alternative decision-support tool for determining the optimal harvesting time in a more objective, consistent, and efficient manner, thereby supporting productivity improvement and reducing the potential for human error during harvest assessment.