Classification and Identification of Weeds Using Machine Learning Classifiers

Abstract

Weeds pose significant challenges in agriculture, impacting crop yields and increasing the reliance on herbicides. Accurate and timely identification of weeds is crucial for effective weed management strategies. In this study, we propose a novel approach for automated identification of weeds using various machine learning classifiers. Our study explores the effectiveness of diverse algorithms, including Support Vector Machine (SVM), Random Forest, Decision Tree, k-Nearest Neighbors (KNN) , Extra Tree, and Gaussian Naive Bayes (NB). By pre-processing and engineering features from a diverse dataset of weed images, we ensure optimal model performance. Through rigorous experimentation and evaluation, we assess the performance of each classifier in weed identification. Notably, the Extra Tree classifier achieves an impressive accuracy of 96.35% and an outstanding kappa coefficient of 96.21%. These findings offer valuable insights into the effectiveness of different classifiers and their potential applications in precision agriculture for targeted weed management and crop optimization