Image detection and classification of oil palm fruit bunches

Abstract

Many computer vision approaches, such as deep learning and machine learning, can be employed in agriculture in this era of artificial intelligence (AI). These computer vision techniques are frequently utilized in product categorization, identification, and estimation. Implementing computer vision in agriculture will boost agricultural output and quality by assisting farmers in monitoring agricultural activities. Currently, the oil palm harvester judges oil palm maturity manually using natural signs like oil palm colour appearance and the number of oil palm loose fruit drops under the tree. This paper focuses on developing automated detection systems using a deep learning model to detect and classify oil palm fruit bunch based on their ripeness level. The ripeness of oil palm can be classified into four maturity levels: unripe, under ripe, ripe, and over ripe. There are two approaches in this paper namely detection of oil palm fruit bunches and classification of oil palm fruit bunches. This paper employes several types of YOLO algorithms such as YOLOv3, YOLOv3-Tiny, YOLOv4, and YOLOv4-Tiny to compare the performance and accuracy of different versions of YOLO. It is shown that YOLOv4 has a higher accuracy of 98.70% in detecting and classifying oil fruit bunches based on their ripeness level compared to YOLOv3, YOLOv3-Tiny, and YOLOv4-Tiny.