School of Information Technology, Engineering, Mathematics and Physics

Drone-Based AI Monitoring of Coconut Palm Beetle Infestation

A Cutting-Edge USP Research Project in Sustainable Agriculture - (2024 – 2025)

Drone-Based AI Monitoring of Coconut Palm Beetle Infestation

A Cutting-Edge USP Research Project in Sustainable Agriculture

The University of the South Pacific is proud to lead a forward-thinking research initiative titled “Drone-Based Monitoring of Coconut Palm Beetle Infestation Using AI.” This project addresses the urgent need for advanced pest monitoring technologies in Pacific Island agriculture, especially the growing threat from the Coconut Rhinoceros Beetle (Oryctes rhinoceros) — a pest capable of devastating coconut plantations and local economies.

🌴 The Challenge

Coconut palms are vital to the livelihood, economy, and environment of many Pacific Island nations. However, traditional beetle monitoring methods such as ground inspections and manual surveys are labor-intensive, slow, and spatially limited, which makes early detection and effective intervention difficult.

💡 The Innovation

Led by Dr. Rahul Kumar and a multidisciplinary team of researchers and students from the University of the South Pacific (USP) and international partners, the project introduces a drone-based environmental monitoring system powered by Artificial Intelligence (AI). The core technology uses:

  • High-resolution aerial imagery from AUTEL Evo II Dual 640T V3 drones
  • Real-time object detection using YOLOv8 deep learning algorithms
  • GIS mapping for visualizing infestation spread
  • Python and MATLAB for data analysis and AI model training

🔍 How It Works

  1. Drone Flights capture aerial images of coconut plantations.
  2. AI Image Recognition identifies signs of beetle damage like V-shaped leaf cuts and boreholes.
  3. Geotagging allows precise mapping of infestation zones.
  4. Model Evaluation uses accuracy metrics like mAP, Precision, Recall, and F1 score to continuously improve results.
  5. Visual Outputs (heatmaps, bounding boxes) assist in real-time monitoring and targeted pest control actions.

📊 Progress Highlights

  • Trained on over 2,500 annotated images (healthy vs. infested trees)
  • Achieved precision of 74.4%, recall of 71.2%, and mAP@50 of 79.9%
  • Integrated Grad-CAM, Sobel Edge Detection, and Histogram Features to improve classification
  • Developed interactive GIS maps to help local farmers and agricultural authorities plan targeted interventions

This project supports:

  • Sustainable pest control strategies with reduced pesticide use
  • Early warning systems for rapid response
  • Data-driven agricultural policy
  • Hands-on student learning in AI, machine learning, and environmental monitoring

For media inquiries, collaborations, or more details, contact
📧 rahul.kumar@usp.ac.fj

 

With a total budget of €27,500, the project is supported by AUF and USP, covering equipment procurement, model development, training workshops, and field deployment.

  • Expand drone flight missions in beetle-prone areas
  • Engage with local farmers and stakeholders through workshops and training
  • Improve model performance under various real-world conditions
  • Release open-access tools and datasets to benefit Pacific Island agriculture