Oil Palm Fruit Ripeness Detection using Deep Learning

Abstract

To detect the maturity level of oil palm Fresh Fruit Bunches (FFB) generally seen from the loose fruit that fell to the ground. This method is always used when harvesting swit coconuts. Even though this method is not always valid, because many factors cause the fruit to fall from the bunch. The manual harvesting process can result in the quality of palm oil being not optimal. For this reason, technology is needed that can ensure the maturity level of oil palm FFB. This study aims to detect the maturity of oil palm FFB based on digital images by applying a deep learning algorithm so that the maturity level can be classified into three categories, namely: raw, ripe, and rotten. The deep learning algorithm was chosen because there have been many studies that have proven its high level of accuracy. This research method starts from; preparation of data, designing architectural models and convolutional neural network parameters, testing models, testing images, and analyzing results. From the results of the study, it was found that the convolutional neural network algorithm can be applied to detect the maturity level of oil palm FFB with an accuracy value of 92% for test data, and 76% for model testing.