Classification of RSS (Ribbed Smoke Sheet) Based on Presence or Absence of Fungus using the NN (Neural Network) Perceptron Method

Abstract

Indonesia is an exporter of natural rubber. One type of processed rubber used as export material is a sheet of smoked rubber or Ribbed Smoked Sheets (RSS) rubber. The quality of Ribbed Smoked Sheets greatly affects the increase in rubber exports. The quality of Ribbed Smoked Sheets has been stipulated in SNI 06-001-1987 and the International Standards of Quality and Packing for Natural Rubber Grades (The Green Book). The process of determining the quality of Ribbed Smoked Sheets is also called the sorting process. However, in some rubber plantations, the process of sorting is still done manually by observing the presence or absence of mold on the surface of the Ribbed Smoked Sheets in plain sight so as to produce inaccurate and subjective qualities. Therefore, this study is intended to carry out a classification process based on the presence of mold on the Ribbed Smoked Sheets automatically. This research uses image processing with rubber sheet image as input and classification results as output. The classification process of Ribbed Smoked Sheets uses the Neural Network Perceptron method with two classifications, namely moldy Ribbed Smoked Sheets and non-moldy Ribbed Smoked Sheets. This study uses 1000 pieces of Ribbed Smoked Sheets images as training data and 50 images of smoke rubber sheet as test data, each test has 100 pieces with an accuracy value of 96% with the best epoch value on the 4th epoch