Automatic Classification of Medicinal Plant Species Based on Color and Texture Features

Abstract

Plants play an important role in nature, providing sustenance, oxygen and shelter for animals. Many plant and herb species have a long history of use as sources for medicines and medical treatments in a field known as traditional medicine. Traditional medicine is still adopted in many countries (such as India and Brazil) for a variety of reasons: plants and herbs are cheap in comparison to other kinds of pharmaceutical drugs, plants are nontoxic and do not impact any side effect when properly used. But the correct identification of medicinal plant species is still a challenging task in machine learning and computer vision. Many automatic systems for plant species recognition have been proposed on the past few years, but most of the proposed systems only present acceptable accuracies for a specific and limited set of plant species. In this work, we develop a new medicinal plant data set based on the extraction of texture and color features from plant leaf images. A complete automatic plant recognition system is proposed, and five well-known machine learning classifiers are tested as the recognition module. Experimental results showed that the best classifiers are able to obtain average accuracies over 97% on the proposed data set.