Nematode detection and Management System in Rice Plant using Image Processing and Machine Learning

Abstract

As Sri Lanka’s dominant staple and major harvested crop, rice supports rural livelihoods while serving as a critical component of national food security and economic resilience. However, rice production is increasingly threatened by plant-parasitic nematodes, particularly root-knot nematodes (Meloidogyne spp.), which cause severe damage to the root system. Early-stage detection is challenging because nematode activity occurs beneath the soil surface, and visible symptoms such as stunted growth and yellowing leaves often appear only after substantial root damage has occurred. Existing agricultural detection tools lack the capability to diagnose root-specific diseases, creating a critical gap in early and accurate nematode identification.

This project proposes a specialized rice nematode detection system that utilizes image processing and machine learning to analyze rice root images and accurately detect nematode infestations. The system integrates two machine learning models: a preliminary root classifier that verifies whether the uploaded image is a rice root, and a nematode detection model that identifies infection severity. By automating this diagnostic process, the system minimizes the need for expert intervention and enables farmers to detect infections at earlier stages.

In addition to detection, the application provides treatment and management recommendations, guiding farmers on appropriate control methods tailored to the type and severity of infestation. Designed with a user-friendly mobile interface, the solution aims to be accessible to Sri Lankan farmers, agricultural officers, and researchers.

Overall, this project bridges a critical technological gap by introducing a practical, low-cost digital tool for enhancing rice crop health and supporting sustainable agricultural practices in Sri Lanka.