Deep Learning Approach for Palm Oil Fresh Fruit Bunches Harvest Decision

Abstract

The efficiency of palm oil harvesting is crucial to ensuring optimal yield and quality of fresh fruit bunches (FFB). Traditional manual harvesting methods often result in inconsistent outcomes due to human error and subjectivity in ripeness evaluation. This study proposes an intelligent, image-based harvesting decision system that utilizes Convolutional Neural Networks (CNN) and Support Vector Machines (SVM) to automate the classification of palm oil FFB ripeness. High-resolution images of palm fruit are processed using Python-based frameworks (Google Colab 3.10.12, YOLOv8) to extract features such as color and texture, which are then used to train the CNN and SVM models. The system architecture includes stages for image acquisition, preprocessing, feature extraction, classification, and decision-making. Both CNN and SVM were evaluated for performance using accuracy, precision, recall, and F1-score. The experimental results demonstrated high classification accuracy, with CNN achieving an average of 0.97 and the highest result recorded at 0.89. The system significantly enhances harvesting decision accuracy and reduces dependence on manual inspection. This study demonstrates the viability of using deep learning and machine learning algorithms for real-time agricultural decision-making. The integration of CNN and SVM not only improves productivity but also contributes to sustainable practices by reducing waste and labor intensity. The proposed system offers a scalable solution that can be adapted for broader smart farming applications, supporting national goals of digital transformation and energy efficiency in agriculture.