AI assisted optimization of water consumption in grapevines

The project addresses water scarcity in agriculture, focusing on vineyards, by developing innovative strategies and technologies to optimize water use. It aims to create advanced water consumption models and a low-cost, handheld 3D imaging system that integrates multimodal data (e.g., thermal, RGB, and multispectral imaging) to measure key plant traits like leaf area, grape cluster weight, and water consumption. Leveraging cutting-edge machine learning techniques, such as zero-shot learning and domain adaptation, the system will enable precise, non-destructive, and dynamic measurements directly in the field. This innovation combines 3D reconstruction with multimodal data to improve water management models, reduce irrigation reliance, and promote sustainable viticulture practices. The project also investigates the impact of canopy management on water use and yield, with potential applications to other crops like tomatoes and orchard trees. Current water budget models rely on climatic variables and static crop coefficients, failing to account for site-specific conditions, dynamic canopy changes, and agronomic practices. Existing methods for estimating leaf area are labor-intensive, destructive, and lack the resolution needed for dynamic modeling. While 3D reconstruction technologies (e.g., LiDAR, stereo cameras) show promise, they are expensive, complex, and sensitive to environmental conditions. Additionally, integrating multimodal data into 3D models is challenging due to issues like image registration and parallax effects. This project addresses these gaps by developing a scalable, multimodal 3D canopy reconstruction system that integrates plant-specific traits into water consumption models, enabling more accurate and sustainable water management in vineyards and other agricultural systems.