Segmentasi Citra Tanaman Obat dengan metode K-Means dan Otsu

Abstract

Segmentation is the most important thing in the object identification process. Because machine learning-based interest segmentation of true color images is the most difficult task in computer vision. Because in the segmentation process there is a separation between foreground and background from a 3 layer RGB image to a layer 1 process to get a complete image without noise, this greatly affects the level of accuracy in image identification. In addition, we use several image processing operators such as filters, holes and openarea to remove image areas that we do not need. Therefore, in this study, we tested the images on 5 types of medicinal flowers using k-means segmentation with values of k=2 and k=3, as well as the otsu method. Both methods of segmentation are carried out by each method to get the appropriate pattern. The goal is to get the important areas that can be calculated by the image identification algorithm. This research uses 250 images and produces 750 patterns for the identification process. The results obtained are 96% to identify the flower type taraxacum laeticolor Dahlst with the K-means k=2 segmentation method.