Robotic Multi-View Framework for Tomato Plant Monitoring in Hydroponic Greenhouses

This paper presents the development and integration of a cost-effective multi-view RGB-D device for continuous, non-destructive monitoring of tomato plants in a hydroponic greenhouse. The system features three RGB-D sensors, arranged vertically to capture the full vertical profile of each plant, and a zero-shot data processing framework integrating visual and depth cues to extract plant growth parameters and fruit ripeness information. An experimental trial conducted in a greenhouse in southern Italy (Oct 2025-Jan 2026) validated the approach against traditional measurement procedures. The results demonstrate a strong correspondence between automated and reference measurements with a Pearson correlation coefficient r > 0.9 for key traits. Regression models using vision-based plant height estimate as biomass predictor were also trained, showing acceptable predictive performance for stem and leaf biomass (normalized RMSE 14–22%). In addition, a comparative analysis of two nutrient management strategies, i.e., traditional periodic solution renewal versus sensor-based nutrient dosing, revealed that sensormanaged plants reach significantly enhanced vegetative growth.