A Deep Learning Approach for High Accuracy Weed Detection and Tracking in an Agricultural Environment

Weed management is a critical challenge in agriculture, directly affecting crop yields, sustainability, and economic efficiency. Traditional weed control methods, such as manual removal and widespread herbicide spraying, are labor-intensive, environmentally harmful, and economically unsustainable. This project aims to develop an AI-driven weed detection, segmentation, and tracking system that enhances precision agriculture practices by enabling real-time monitoring and intervention.
The research focuses on integrating deep learning and computer vision techniques to accurately classify weeds and crops, improving decision-making for farmers. The key innovation is the ability to track weed growth dynamically, allowing adaptive control strategies that reduce chemical dependency while optimizing crop health. The system will be trained on diverse real-world agricultural datasets, employing convolutional neural networks (CNNs) for high-accuracy detection and segmentation methods to refine weed localization.
By introducing an intelligent tracking mechanism, the system will continuously monitor weed development, predict regrowth patterns, and enable precise, targeted interventions. This research contributes to sustainable farming by reducing herbicide use, minimizing environmental impact, and enhancing crop productivity. The project’s findings will support the development of scalable, AI-powered solutions for modern agriculture, ultimately improving efficiency and promoting eco-friendly farming practices.

Faculty Supervisor:

Majid Ahmadi

Student:

Partner:

BHF Agrobot

Discipline:

Engineering

Sector:

Agriculture; Manufacturing

University:

University of Windsor

Program: