Thermal imaging and deep learning system for optimizing water and nutrient use in tomato crops

Overview

This research initiative develops a deep learning-driven thermal imaging system designed to improve water and nutrient use efficiency in tomato crops grown under protected cultivation. By modeling heat exchange across the soil-plant-atmosphere system, the platform identifies plant stress conditions and informs smarter fertigation decisions. The solution addresses a critical challenge in Colombian agriculture, where inefficient resource use leads to low yields and reduced competitiveness, while also contributing to broader food security goals across diverse climatic regions.

Technical specifications

• Thermal data acquisition using Short-Wave Infrared (SWIR) cameras combined with temperature sensors in the substrate, plant canopy, and microclimate
• Deep learning algorithms trained to classify and predict thermal behavior across the substrate-plant-microclimate system
• Experimental design comparing tomato crops under free exposure versus two simulated plastic covering models, at 250 m² plots with commercial planting density
• Multivariate analysis tracking water use efficiency, fertilizer efficiency, and crop yields under each condition
• Statistical validation comparing variables across treatments to confirm model reliability
• Fertigation integration translating thermal imaging outputs into automated irrigation and nutrition control

Technology readiness level

The project is in the early experimental phase. Initial data collection and model development will take place in research centers, with subsequent testing under local grower conditions over a two-year period. The research is conducted jointly with producers and university researchers, with future plans to expand collaboration through cloud-based data sharing. Hardware and software refinements based on field results will be pursued to move toward a commercially integrated package. Additional regional researchers will be engaged to strengthen the algorithms and broaden applicability.