A tool developed at the University of Bonn will be able to predict crop yields and other functions to help farmers, according to a study published in Plant Methods.
A team of researchers from the University of Bonn, Germany, developed AI software to simulate the growth of different field crops. They fed an AI system thousands of photos from field experiments. Using these images, the algorithm “learnt” to predict how plants develop based on a single image during the initial stages of growth. As a result, the system can generate AI images showing the future development of the crop and calculate parameters such as leaf area and corp yields accurately. “We have developed software that uses drone photos to visualize the future development of the plants shown,” explained Lukas Drees from the Institute of Geodesy and Geoinformation at the University of Bonn. “We took thousands of images over one growth period. In this way, for example, we documented the development of cauliflower crops under certain conditions.”
The whole process is highly accurate, but the crop conditions must remain predictably similar to those present when the training photos were taken. For example, the software can’t consider a sudden cold snap or steady rain lasting several days. However, the authors expect that the system will eventually be able to learn how crops are affected by events such as these, and predict an increased need for fertiliser, for example. In the future, farmers will be able to increasingly rely on computer and AI systems to manage their farms, including when to use pesticides and fertilisers
“In addition, we used a second AI software that can estimate various parameters from plant photos, such as crop yield,” said Drees. “This also works with the generated images. It is thus possible to estimate quite precisely the subsequent size of the cauliflower heads at a very early stage in the growth period.”
One area of focus is the use of polycultures, which means farmers sow more than one species in the same field, such as beans and wheat, for example. As plants have different requirements, they compete less with each other compared to a monoculture, where just one species is grown. This boosts yield and may reduce the need for fertiliser. For example, some species – such as beans – can bind nitrogen from the air and use it as a natural fertiliser. Inevitably, the second species (wheat in this example) also benefit from this.
“Polycultures are also less susceptible to pests and other environmental influences,” explained Drees. “However, how well the whole thing works very much depends on the combined species and their mixing ratio.” When results from many different mixing experiments are fed into learning algorithms, the AI system can learn which plants are particularly compatible and in what ratio.
Computer simulations, up until now, have tried to understand what nutrients and environmental conditions crops need during their growth in order to thrive. “Our software, however, makes its statements solely based on the experience they have collected using the training images,” stressed Drees.
Both approaches should complement each other. The authors believe that if both methods are combined, they could significantly improve the quality of the forecasts. “This is also a point that we are investigating in our study,” concluded Drees. “How can we use process- and image-based methods so they benefit from each other in the best possible way?”
Drees, L., Demie, D.T., Paul, M.R. et al. Data-driven crop growth simulation on time-varying generated images using multi-conditional generative adversarial networks. Plant Methods 20, 93 (2024). https://doi.org/10.1186/s13007-024-01205-3