Plant Leaf Recognition: Comparing Contour-Based and Region-Based Feature Extraction

Abstract

Plants play a vital role in the environment. Identifying them and classifying them is an important task for botanists. This study briefly points out- how to recognize plant species using image processing techniques that can help botanists and scientists, the appropriate features for plant species recognition in feature extraction, how can a classification help to increase the accuracy of the plant leaf classification. There are four major phases used in here for the recognition, and they are image input, image pre-processing, feature extraction, and SVM classification. This automatic recognition system is developed using python with Jupyter Notebook environment gives higher accuracy for the plant recognition for the botanists and comparing the feature extractions such as Contour-based and Region-based to get down more accurate results than previous researches is the main purpose of the proposed study. Contour-based and Region-based features were calculated through equations. SVM classification is used for both feature extraction methods. For individual feature extraction the Contour-based feature extraction is more efficient with 72.25% accuracy than Region-based feature extraction with 70.41% accuracy, and for combining both feature extraction SVM classification gives 68.58% accuracy. Contour-based feature is the most appropriate feature for a plant species recognition.