Penggabungan Dimensi Fraktal dan Kode Fraktal untuk Identifikasi Tumbuhan Obat Indonesia Menggunakan Probabilistic Neural Network

Abstract

This research proposed a new method to identify Indonesia medicinal plants using fractal dimension (FD) and fractal code (FC) using Probabilistic Neural Network (PNN). This research investigates the effectiveness of fusion of FD and FC in order to identify Indonesia medicinal plants. FD is measured by Box Counting method. FD has non-integer value that represents the self-similarity of fractal. FC is based on the self-similarity in a picture. It means that small pieces of the picture can be approximated by transformed versions of some other (larger) pieces of the picture. The fusion of FD and FC is done by using vector fusion and Product Decision Rule (PDR). The total medicinal plants used in this research are 20 species taken from Biofarmaka Farm, Cikabayan and Green House Center Ex-Situ Conservation of Medicinal Plants Indonesia Tropical Forest. Each species consists of 30 images, thus the total images used in this research are 600 images. The PNN is trained using 450 images to classify 20 kinds of plants. The experimental results shows that fractal dimension has an accuracy rate of 57%, fractal code 21%, vector fusion 58%, and PDR 58%. The experimental result also shows that fractal fusion is not necessary because the fractal dimension has significant information compared to the fractal code.