Edge-Deployed Visual Pest Detection System For Real-Time Crop Protection

Abstract

Early and precise identification of agricultural pests is essential to avoid crop damage and maximize agricultural output. An edge-deployable pest detection system based on convolutional neural networks (CNNs) and deep learning models for real-time monitoring and pest infestation early detection is the premise of this research. Natively designed for deployment in field conditions, the system runs on low-power edge hardware like Raspberry Pi and NVIDIA Jetson Nano, carrying out on-site image processing independent of cloud connectivity. Embedded camera-captured images of crop leaves are processed locally to identify and classify pest species. Moreover, the system includes an alert mechanism in real-time through visual displays, or smartphone push notifications to alert farmers in due time, thus minimizing yield loss. Systematic tests on public and curated pest image datasets showed high precision, low latency, and effectiveness in terms of resource utilization. This study highlights the opportunity of integrating deep learning and edge computing technologies toward intelligent agriculture, especially for remote or resource-poor locations.