Automated Classification of Refined, Bleached and Deodorised (RBD) Palm Olein Grades CP6 and CP8 Using Deep Learning and Computer Vision

Abstract

This study investigates the use of deep learning and computer vision to automate the classification of Refined, Bleached, and Deodorised (RBD) palm olein grades, specifically for the CP6 and CP8 categories. The primary challenge in the current industry is the reliance on manual inspection, which is prone to human error, and chemical analysis, which is both destructive and costly. Therefore, the main objective of this research is to develop a multi-class Convolutional Neural Network (CNN) model to automatically differentiate between these oil grades. The methodology involves developing a CNN model using the TensorFlow framework within a Google Colab environment, with training limited to five epochs to ensure compatibility with industrial edge devices. Furthermore, Fourier-Transform Infrared (FTIR) spectroscopy was utilised as a validation method to identify molecular differences between the two grades. The findings demonstrate that the CNN model achieved promising performance, with precision exceeding 0.70 and recall values above 0.62. FTIR analysis confirmed significant differences in carbonyl (C=O) stretching, which scientifically supports the visual-based classification. In conclusion, the integration of Artificial Intelligence into palm oil grading provides a rapid, non-destructive, and efficient alternative, supporting digital transformation in the palm oil sector in alignment with the goals of Industrial Revolution 4.0.