(Utilization of Computer Technology for Medicinal Plant Identification Based on Leaf Image)

Yeni Herdiyeni1*, Julio Adisantoso1, Ellyn K Damayanti2, Ervizal AM Zuhud2,
Elvira Nurfadhila1, Kristina Paskianti1

ABSTRAK

Indonesia adalah negara yang kaya akan keanekaragaman tumbuhan, termasuk tumbuhan obatnya. Dengan banyaknya spesies dan jumlah tumbuhan obat di Indonesia maka identifikasi tumbuhan tidaklah mudah. Oleh karena itu, diperlukan sistem komputer yang dapat membantu masyarakat mengidentifikasi tumbuhan dengan mudah. Ada sebanyak 30 spesies tumbuhan obat yang digunakan dalam penelitian ini. Penelitian ini menggunakan 3 penciri (fitur) tumbuhan obat, yaitu morfologi, tekstur, dan bentuk. Untuk mengelompokkan spesies tumbuhan obat digunakan probabilistic neural network. Hasil percobaan menunjukkan akurasi identifikasi mencapai 74,67%. Pengembangan sistem mesin pencari (search engine) untuk tumbuhan obat juga sudah dilakukan. Sebanyak 32 jenis dokumen tumbuhan obat digunakan dalam sistem ini dengan jumlah dokumen sebanyak 132 dokumen. Setiap dokumen terdiri atas nama, famili, deskripsi, penyakit, dan kandungan kimia. Pengelompokkan dokumen penelitian ini menggunakan KNN Fuzzy. Dokumen tumbuhan obat dibagi menjadi 2 kelompok, yaitu berdasarkan famili dan penyakit. Hasil percobaan menunjukkan akurasi rata-rata precisi untuk mesin pencari mencapai 96% untuk kueri dengan panjang 1 kata dan 89% untuk kueri dengan panjang 2 kata. Sistem ini sangat bermanfaat bagi masyarakat untuk membantu mengidentifikasi tumbuhan obat dengan mudah sehingga masyarakat mampu memanfaatkan tumbuhan obat guna mendukung kegiatan keanekaragaman tumbuhan di Indonesia.

Kata kunci: bentuk tanaman obat, identifikasi tumbuhan obat, morfologi tanaman obat, probabilistic neural network product decision rule

ABSTRACT

Indonesia is a mega biodiversity country including many kind medicinal plants. It is not easy to identify the various kinds of the medicinal plants especially for common people. Therefore, we need a computer-based automatic system as a tool to identify these various types of the medicinal plants. Developing of computer-based automatic system for medicinal plant identification has been done based on leaf image. There are 30 species of medicinal plants used in this study. There are 3 features for identification, i.e. morphology, texture, and shape. To improve the accuracy of identification we applied probabilistic neural network to classify the species of medicinal plant. The experiment results showed that the accuracy of identification increase to 74.67%. Developing of search engine has been done as well. We used 32 species of medicinal plant. The number of document was 132 documents. The document consists of name, family, description, diseases, and chemical substances. To improve the accuracy of searching, we applied KNN Fuzzy to classify document into 2 categories, i.e., family and diseases. The experiment results showed that the accuracy of average of precision is 96% for only word of length query and 89% for two words of length query. The system is very beneficial for people in society because it can be used to identify medicinal plants easily and the relevant communitis become independent in maintaining family health and giving opportunities as well as income of the people. Hence, the system is promising for leaf identification and supporting plant biodiversity in Indonesia.

Keywords: leaf identification, leaf morphology, leaf shape, medicinal plants, probabilistic neural network, product decision rule