Term 252

PalmGuard: Red Palm Weevil Detection System

Project Type: Self-Initiated

Project Description

Red Palm Weevil (RPW) infestation remains a critical threat to date palm plantations, particularly in large-scale agricultural regions such as Saudi Arabia, where early detection is challenging due to the pest’s hidden internal activity. This project presents PalmGuard, a distributed smart monitoring system designed to enable early detection of RPW infestation and improve inspection efficiency at the plantation scale. The system integrates low-power embedded sensor nodes, wireless communication, cloud-based machine learning, and a mobile application into a unified architecture. Each sensor node, built around an ESP32 microcontroller and a vibration sensor, is attached to palm trunks to continuously collect vibration signals associated with potential larval activity. The collected data is transmitted via a LoRa network to a central gateway, which forwards it to a cloud backend. A deep learning model processes the vibration data and classifies infestation likelihood, generating confidence-based alerts. These alerts are then delivered to users through a cross-platform mobile application that provides real-time notifications, palm-level status, and inspection guidance. To improve operational efficiency, the system incorporates an optimization module based on the Team Orienteering Problem (TOP), which prioritizes inspection routes based on risk scores derived from machine learning outputs, palm characteristics, and spatial factors. The prototype was implemented and tested under controlled conditions. Results show that the communication subsystem achieved approximately 100% packet delivery over 60 meters, and the sensor node operated within low-power constraints, supporting long-term deployment. The machine learning model achieved an alert validity rate of approximately 93.45% and inference latency well below the required threshold. Backend performance testing confirmed that API response times met the specified P95 ≤ 500 ms requirement, and end-to-end alert delivery was achieved within 6–9 seconds. The optimization module demonstrated significant improvements in inspection efficiency, achieving over 60% risk coverage within daily constraints and reducing inspection labor requirements by approximately 50% compared to baseline manual methods. These results indicate that PalmGuard provides an effective and scalable solution for early RPW detection and inspection optimization. Future work includes extended field validation under real environmental conditions, expansion of the dataset to improve model robustness, and further system optimization for large-scale deployment.