Herbal Leaf Pattern Analisis Using Principal Component Analisis (PCA) and Canberra Distance

Herbal plant material is very much encountered in various regions so it is very easy to find. Various herbs can be used as medicine but because of the limitation to recognize the characteristic of herbal leaves then the user of the herb plant as the drug is still lacking. In this study, developed an identification system of herbs based on leaf pattern. Before identification, pre-processed stages should be performed first such as RGB image conversion to grayscale, image segmentation, histogram equalization, and resizing of herbal leaf images. The feature extraction method used in this system is the method of Principal Component Analysis and Canberra distance. The research was conducted on 40 types of herbal leaf. Based on research, the percentage of recognition in the herbal leaf pattern recognition system used Canberra's distance with 100 main components reaching 95.83%, 75 main components reached 95%, and the main component 50 reached 94.16%. Testing of the imagery outside the database using 10 leaf images and 100 main components resulted in a recognition percentage of 70%.