Tea Plant Health Monitoring and Treatment System for Enhanced Sustainability

Abstract

Sri Lanka's tea industry plays a major role in the national economy. Known for hosting prestigious events like the Commonwealth and Olympic Games, Sri Lanka holds the distinction of producing the world's first "Ozone Friendly Tea" as recognized under the Montreal Protocol Agreement. Despite its accolades, the industry faces significant challenges, particularly with regard to the health and productivity of tea plants. These challenges are exacerbated by reliance on expert evaluations and resulting delays in treatment, which impairs farmers' ability to effectively manage plant health for reduced productivity and reduced quality leaves. This study introduces a new approach designed to assist farmers in monitoring and managing the health of their tea plants, thereby increasing productivity and leaf quality. Our solution features a user-friendly interface that enables farmers to upload images of their tea plants. The system analyzes these images and provides feedback on the health status of the plants with recommendations for appropriate treatments or interventions. To refine the accuracy, the uploaded images are classified into relevant classes in the dataset with the help of domain experts. The system democratizes access to expert-level knowledge, empowering farmers with the information they need to improve the health, productivity, and quality of their tea plants. For the detection and prevention of tea leaf diseases, image processing techniques and convolutional neural networks (CNN) were used. Initially, the authors combined two datasets, which were then shown to the Tea Research Institute (TRI) to get feedback on the dataset. Images were subjected to preprocessing, which included data augmentation and resizing before being fed to the network. To minimize overfitting and increase CNN accuracy, iterations were adjusted as needed. The models achieved an outstanding accuracy of 94.80%. By enlarging the dataset, this CNN-based method improves the classification and detection performance of tea leaf diseases, providing significant improvement over traditional methods.