Identification of Herbal Leaf Types Based on Their Image Using First Order Feature Extraction and Multiclass SVM Algorithm

Abstract

One way to increase immunity and maintain immunity can be done by consuming herbal plants. This herbal medicine is empirically believed to be useful as a cultural treasure from generation to generation. All parts of the plant can be used as medicine, one of which is the leaves. However, most people do not know the herbal leaves. This herbal leaf can actually be recognized from the characteristics of its shape. This study aims to identify types of herbal leaves using first-order feature extraction and the Multiclass Support Vector Machine (Multiclass SVM) algorithm. First-order feature extraction is able to extract features using the parameters of mean, skewness, variance, kurtosis, and entropy. Meanwhile, Multiclass SVM identifies by obtaining the optimal line in separating the data set of two classes of two-dimensional space points in order to find the maximum hyperplane in separating the data points into classes so that they can be grouped. From the test results, the identification accuracy is an average of 76%. This shows that the algorithm has been able to identify, but needs improvement.