Feature Selection Application to Classify Medicinal Plant Leaves using LVQ

Abstract

Indonesia is a country famous for cultivating medicinal plants, but the large number of medicinal plants is inversely proportional to the number of people who have the knowledge and can recognize the types of medicinal plants. To solve this problem, a computer vision system was developed that aims to identify medicinal plants based on the image of their leaves. Several stages used in this research are image preprocessing, image segmentation, feature extraction, feature selection and classification using the Learning Vector Quantization (LVQ) method. The optimal features used are texture features B (ASM 0°, Contrast 0°, Contrast 45° and Contrast 90°) and a combination of shape and texture features B (Diameter, ASM 0°, Contrast 0°, Contrast 45° and Contrast 90°). The optimal parameter used in the LVQ method is the learning rate (a) which is 0.1, the decrease in the learning rate (dec a) is 0.75. When using the texture B feature, the training accuracy is 95.52% and the testing accuracy is 95%. When using a combination of shape and texture B features, the training accuracy is 94.40% and the testing accuracy is 92.50%.