Identification of Zingiberaceae Species Based on Leaf Recognition Using Multiclass Support Vector Machine

Abstract

The family Zingiberaceae or known as the ginger family of flowering plants famous for its medicinal values and is widely distributed especially in Southeast Asia. Malaysia is one of the countries with abundance of Zingiberaceous species. Usually, identification of the plant species begins by identifying the leaves. Identification based on leaf recognition is the most effective method because a healthy plant has leaves and it exists all the time, unlike fruits and flowers which may only exist at certain times. However, the leaf recognition considered as intricate task and challenging especially when using conventional approaches as most plants have similar shape and colour. Thus, the aim of this research is to develop an interactive application for classification and identification of selected Zingiberaceae species, namely: Zingiber officinale, Curcuma longa, Etlingera elatior and Alpinia galanga based on leaf recognition using multiclass Support Vector Machine (SVM). Six steps were constructed to develop an interactive application for classification of Zingiberaceae species: 1) Collection of leaf samples, 2) Creation of database from captured leaf images, 3) Image preprocessing, 4) Leaf image processing, 5) Classification of image textures using SVM and 6) Design of graphical user interface (GUI). The image features extraction in leaf image processing were based on gray-level co-ccurrence matrix (GLCM), Canny and Prewitt algorithms. The combination of GLCM and Prewitt achieved the highest accuracy which was recorded of 95% from the overall accuracy in classification of Zingiberaceae species. By using the automatic plant identification system, results will come up more accurate and faster.