Walk through a lettuce field and it’s easy to assume every plant is growing the same way.
In reality, no two plants are identical.
Some develop faster than others. Some receive more nutrients. Some begin showing signs of stress long before they’re visible to the human eye. Yet farmers often have to make decisions—when to fertilise, irrigate or harvest—based on measurements taken from only a small number of plants.
The rest of the field is, quite literally, an educated guess.
Researchers from the National Research Council of Italy (CNR), the University of Bari and the AgRibot project are working to change that.
Their latest research introduces an automated monitoring system capable of analysing individual lettuce plants as a robotic platform moves through the field, creating a detailed picture of crop growth without the need for hours of manual measurements.
Seeing what the human eye can’t measure
Traditionally, assessing crop growth means stopping in the field with rulers, specialised instruments and notebooks.
Plant height, leaf area, biomass and crop health are measured manually before the data is used to guide management decisions. While these methods are reliable, they are also slow, labour-intensive and only represent a fraction of the plants growing in the field.
The system developed by the researchers approaches the problem differently.
Mounted on a mobile robotic platform, it combines RGB cameras, depth sensors, infrared imaging and high-precision GPS to capture a detailed digital representation of every plant it passes. Instead of simply taking photographs, the platform reconstructs each lettuce plant in three dimensions, allowing it to estimate characteristics such as height, diameter, projected leaf area, plant volume and vegetation health automatically.
Even more importantly, the researchers designed the system to work with minimal human input by using zero-shot AI models that recognise lettuce plants without requiring manually labelled training datasets for every new field.
Technology is only useful if the information it produces can be trusted.
To validate the system, the researchers monitored lettuce crops throughout an entire growing season in Bari, Italy. The robot’s measurements were compared with traditional field measurements and laboratory analyses.
The results showed a strong agreement between the automated estimates and conventional methods, with correlation coefficients above 0.9 for key plant traits. In other words, the robotic platform was able to monitor crop development with a level of accuracy that closely matched manual assessments.
The study also demonstrated that the system could detect differences in crop development under varying nitrogen fertilisation levels, providing valuable information that could help growers optimise fertiliser use while reducing unnecessary inputs.
Precision agriculture has always relied on one essential ingredient: information.
The better farmers understand how crops are growing, the more precisely they can manage water, fertilisers and other resources. But collecting that information across an entire field has traditionally required significant time and labour.
This research brings agriculture a step closer to continuous, automated crop monitoring, where robots don’t simply carry out tasks—they also observe, measure and interpret what is happening in the field as crops develop.
Rather than replacing agronomic expertise, systems like these provide farmers with richer, more frequent information, allowing decisions to be based on the condition of individual plants instead of field-wide averages.
As agricultural robotics continues to evolve, the ability to understand crops plant by plant may become just as valuable as the ability to work the field itself.