Drone-based Pest Control: LiteDepthwiseNet Stem Borer Detection on Jetson Nano

Excessive use of pesticides in agriculture has led to widespread environmental contamination, pest resistance, and unwarranted chemical exposure. This paper presents a drone-based precision pest control system that detects and eliminates stem borers in specific regions based on multispectral imaging. A drone-borne multispectral camera captures spectral data, which is processed in real time using LiteDepthwiseNet, an efficient resource deep-learning framework for multispectral image classification. LiteDepthwiseNet, with its 3D depthwise convolution and focal loss optimization, can correctly detect pests in a computationally efficient manner that enables deployment on edge devices like the Jetson Nano. Onboard Jetson Nano processes multispectral images for stem borer identification and accurately activates a targeted elimination mechanism with minimal pesticide application to infested areas. The drone is controlled from the ground to facilitate adaptive, site-specific treatment without spraying large-scale pesticides. Compared with conventional pest management practice, the system significantly reduces pesticide waste, environmental pollution, and enhances agricultural sustainability. With field experiments, the system is demonstrated to have high detection accuracy and real-time inference with almost zero latency, and therefore an efficient and scalable solution for precision agriculture. By integrating multispectral image analysis with edge AI processing, this study presents a novel, low-cost pest control solution, opening the door to more productive and sustainable farming. The solution demonstrates the viability of AI-driven, drone-based pest control as an ideal substitute for chemical-based alternatives, ensuring higher agricultural output with less environmental harm.