Enhancing Plant Leaf Identification: A Comparative Study of Machine Learning Models

Abstract

In this paper, we present a comprehensive analysis of the application of various machine learning models to the task of plant leaf identification. A Weiner filter is applied for noise reduction in the data, and morphological operations are utilized for feature extraction. Subsequently, evaluated the performance of eight different classification models: Quadratic Discriminant Analysis (QDA), Extra Trees Classifier, Random Forest Classifier, Linear Discriminant Analysis, SGD Classifier, Bagging Classifier, Perceptron, and AdaBoost Classifier. These models are assessed in terms of their accuracy, balanced accuracy, ROC AUC, F1 Score, and time taken for predictions. The results reveal that QDA emerges as the top-performing model, achieving remarkable accuracy and balanced accuracy of 93%, along with an F1 Score of 93%. Extra Trees Classifier and Linear Discriminant Analysis also exhibit strong performance with high accuracy and balanced accuracy scores. The SGD Classifier, Bagging Classifier, and Perceptron yield competitive results as well. However, the AdaBoost Classifier falls short in terms of accuracy and F1 Score, indicating challenges in plant leaf identification. The Random Forest Classifier, while achieving an accuracy of 87%, shows slightly lower balanced accuracy and F1 Score.

Index Terms- Plant leaf identification, Wiener Filter, Extra Trees classifier, Bagging classifier, Balanced accuracy.