Implementation of YOLOv9 in Agricultural AI for Enhanced Weed Detection

Abstract

Weeds significantly hinder agricultural productivity by reducing crop yields and increasing production costs. Leveraging artificial intelligence (AI) is critical for equipping farmers with early detection capabilities to implement effective weed management strategies. Among AI technologies, deep learning (DL) techniques are especially effective for analyzing agricultural field images to identify weed species. This paper reviews state-of-the-art DL methodologies, with a focus on the YOLOv9 model for weed detection. Our findings highlight YOLOv9’s superior effectiveness, achieving an accuracy of $\mathbf{9 0 \%}$, due to its robust architecture and advanced feature extraction. The model’s proficiency in accurately identifying weeds is evident from our thematic analysis. We also explore various image acquisition devices, including robots, drones, and mobile phones. Comparative evaluations show that YOLOv9 consistently outperforms other DL techniques, achieving remarkable accuracy and speed. This study includes a comparative analysis of several algorithms like SSD, Mask R-CNN, and Fast R-CNN, underscoring YOLOv9’s superior performance. This paper serves as a valuable resource for researchers and practitioners, guiding future efforts in weed management and precision agriculture by emphasizing YOLOv9’s transformative potential.