Journal of Agronomy and Field Innovations

Background: Tomato (Solanum lycopersicum L.) is one of the most commercially viable plants, but it is very vulnerable to fungal, bacterial, and viral diseases, which can negatively affect the yield and quality of produce. Traditional scouting relies on visual methods, which need a lot of time and effort. Usually, conventional methods of scouting fail to detect diseases in their early stages.
Objective: The objective of this study was to assess the performance of an integrated digital phenotyping and machine vision framework for rapid, non-destructive and early detection of major tomato diseases in a greenhouse environment.
Methods: An Internet of Things (IoT) enabled environmental monitoring network was integrated with RGB, multispectral, hyperspectral and thermal imaging in a naturally ventilated polyhouse experiment for one cropping cycle. The image datasets were analyzed using a machine vision pipeline consisting of image pre-processing, segmentation, feature extraction, and AI-assisted classification. The framework was tested for early detection of bacterial spot, early blight, late blight, powdery mildew and wilt disease of Fusarium. The imaging modalities were compared in terms of disease incidence, disease severity, plant growth parameters and classification performance indicators including accuracy, precision, recall and F1-score.
Results: The integrated digital phenotyping framework detected disease symptoms 4–7 days ahead of conventional visual scouting. The AI-assisted classification model obtained overall accuracy, precision, recall, and F1-score of 96.4%, 95.8%, 95.1%, and 95.4%, respectively. Hyperspectral imaging was the most sensitive for pre-symptomatic disease stress detection, while RGB imaging was the most cost-effective solution for routine greenhouse monitoring. The integrated system also reduced labor requirements by an estimated 62% compared to manual disease scouting and enhanced diagnostic consistency and reliability.
Conclusion: The integration of multi-modal imaging, artificial intelligence-based machine vision and IoT enabled environmental sensing provides a robust, scalable and non-destructive approach for early disease diagnosis in protected tomato cultivation. The proposed framework has great potential to improve precision disease management, reduce unnecessary application of agrochemicals, improve greenhouse productivity and support sustainable vegetable production systems.