Identifikasi Kecambah Kelapa Sawit Palsu Menggunakan Metode Convolution Neural Network

Abstract

Indonesian Oil Palm Research Institute (IOPRI) is one of the oil palm sprout research institutions in Indonesia to produce superior seeds with international standards. Besides the value of quality seeds, there are also various irresponsible parties out there who want to take advantage of this. . They market fake oil palm sprouts on behalf of PPKS and the same characteristics between original and fake sprouts make many farmers deceived by buying fake sprouts. The use of fake sprouts will also affect the productivity of oil palm plantations in the future. Therefore, a system that is able to identify fake oil palm sprouts is needed. This research will identify fake oil palm sprouts with the Convolutional Neural Network algorithm as a machine learning method and images of oil palm sprouts as a form of input data. The original PPKS oil palm sprout image data was collected as many as 600 images and 510 images of fake oil palm sprouts. So that the total data used amounted to 1110 images. A total of 888 images are used as training and 111 as testing data. Resizing and augmentation techniques are carried out as an image preprocessing stage. The next step used Xception architecture as an identification method followed by several hyperparameters. Based on the results of system testing, an accuracy of 96% is obtained. In addition, the distance also affects the identification results, therefore it is recommended to take pictures no more than 10 cm.