Optimization of Fuzzy Local Binary Pattern in Threshold and Operator Selection using Multi Objective Genetic Algorithm

Abstract

This research proposes multi-objective genetic algorithm non-dominatedsorting (MOGA NSGA-II) of fuzzy local binary pattern to optimize LBP operator and fuzzy threshold for identification of Indonesia medicinal plant. Multiobjective genetic algorithm (MOGA) is genetic algorithm (GA) which developed specifically for problems with multiple objectives. We evaluated 1,440 medicinal plant leaf images which are belonging to 30 species. The images was taken from Biofarmaka IPB, Cikabayan Farm, Green house Center Ex-Situ Conservation of Medicinal Plant Indonesia Tropical Forest and Gunung Leutik. FLBP is used to handle uncertainty on images with various patterns. FLBP approach is based on the assumption that a local image neighbourhood may be characterized by more than a single binary pattern. The experimental results show that the correct selection of FLBP operator and threshold can improve the identification from 66.44% to 85%. It can be concluded that this propose method is capable to improve medicinal plants identification species efficiently and accurately.