MULTI-SPECTRAL EMBEDDED VISION SYSTEM FOR EARLY PLANT DISEASE DETECTION IN PRECISION FARMING

Abstract

Agriculture depends on detecting plant disease early and using pesticides efficiently. Traditional methods, such as manual inspection and spraying large areas, are slow, harmful, and environmentally wasteful. These systems often lack real-time response and precise targeting. This project introduces a multi-spectral embedded vision system designed to revolutionis how plant diseases are detected and treated in precision farming. The traditional methods rely on human inspection and blanket pesticide spraying. This system replaces this with an intelligent and automated solution. By using a RetineXNet to enhance lighting conditions and DnCNN to clean noisy images and the system ensures high-quality input data for disease classification. These AI models run in Python and communicate with an ESP32 microcontroller. Which controls a relay-driven pesticide pump, displays the disease type on an LCD and triggers an alert. They are all powered by a 12V battery and making it fully field deployable. The innovation lies in combining multi-spectral imaging with real-time embedded AI. To enable early detection and targeted pesticide application. This not only improves accuracy but also drastically reduces chemical usage and making farming smarter, cleaner and more sustainable.