IDENTIFICATION OF PATCHOULI LEAVES QUALITY USING SELF ORGANIZING MAPS (SOM) ARTIFICIAL NEURAL NETWORK

Abstract

One of the essential oil export commodities from Indonesia is patchouli oil. However, the price of patchouli todays is unstable caused by the low quality of the oils, which has high levels of acid and lower alcohol content. One part of patchouli that is widely used to obtain essential oils is the leaf. The better quality of leaves will produce oil with grade quality. The quality of the leaves can be identified by its physical characteristics. Leaves that have a good quality are small leaves, thick and slightly yellowish red color. This identification process can be done visually, but, it will be easier if it can be done automatically using computer applications. Therefore, this paper performs automatic identification of leaves utilizing image of patchouli leaves and artificial neural network algorithm Self Organizing Maps (SOM). Identification was done to distinguish the leaves with good quality and poor. From the test results using the initial learning rate 0.1, 0.3 deduction learning rate, the minimum rate learning 0.0001, 40 training data and testing the data 60 obtained an average accuracy of 82.82%.