Starting in 2020, a part of the food industry has been scrambling to find new solutions to extend the shelf life of potatoes. The reason? Beginning that year, first in Europe and then in Switzerland, Chlorprophamium, an insecticide now considered toxic to human health but until then used to prevent premature sprouting in tubers that causes loss of quality, weight and thus economic revenue, was banned.
In response to this challenge, traditional storage practices (cool, dark and ventilated environments between six and eight degrees) have come to the aid of another compound, this time harmless to consumers but with a not entirely negligible flaw: a decidedly high cost.
It is Orange Oil, a natural substance extracted from the orange peel that industries spray on potatoes stored in warehouses, penetrates the tuber skin and prevents sprouting. It works, but it's expensive. So they wanted to figure out how to optimize its use, and apply it only when it's really needed. Spraying should not be done too early, otherwise it will wear off the inhibiting effect, but neither should it be done too late otherwise there is a risk of being unable to stop the germination process.
At first glance, one might think of monitoring the stored potatoes with video cameras to detect the exact germination time and intervene in time. But this is not an effective solution, because the application of Orange Oil must take place at least two weeks before the emergence of roots. After this time, the inhibitory effect of the compound will no longer be effective.
It is then a matter of finding the precise time, and this is where artificial intelligence comes in. The PRONTO (Predicting Potato Sprouting to Optimise Tuber Storage) project, of which 精东影业's Institute for Information Systems and Networking (ISIN) is a member, aims precisely to develop solutions based on predictive models that can detect early when potato sprouting occurs.
Researchers Omran Ayoub and Davide Andreoletti have shared much of their academic careers. First in Milan as fellows in the master's and doctoral programs at the Politecnico and now together at ISIN in the Department of Innovative Technologies, where they work as data scientists and also follow PRONTO closely:
"The results so far are more than encouraging. In the past, predicting the time of potato germination involved a margin of error of 30-40 days, but today, thanks to the new algorithms, this interval has been reduced to just seven days. This is a remarkable progress that makes it possible to halve the applications of Orange Oil, with obvious economic savings, especially for food companies that manage large potato warehouses and with important implications for better waste management".
The two researchers explain the revolutionary vision approach provided by the food companies involved in the project: "it is crucial to understand what is happening inside the plant, and to do this it is necessary to listen to its signals, collecting electrophysiological data and other behavioral indicators that can be exploited by AI algorithms to determine the time of germination a few weeks earlier."
Plants, then, talk, and understanding their language requires going through sensors that can detect and transmit a large amount of data.
"Sensors are inserted into hundreds of potatoes to monitor electrophysiological signals. At the same time, the exact date of germination is recorded. The collected data are then processed by the algorithm, which is also giveen the precise information about the day of germination. Thanks to this combination of data - on the one hand, the electrophysiological signals, on the other hand, the actual dates of germination - the predictive models are asked to provide as accurate an estimate as possible of when germination will occur, and as a result, it will be understood when is the ideal time to apply spraying of Orange Oil"
The project began in July 2023 and will end in the fall of 2025, and over the next year there will be room to refine the data collection technique, which changes in consideration of the different types of potatoes that exist, where they come from, the environment they are grown in (greenhouse or outdoors), and numerous other factors.
PRONTO is funded by Innosuisse and implemented in collaboration with several agricultural and food companies, including Vivent Biosignals, UPL, Fenaco, and Zweifel. The federal competence center Agroscope and the Fernfachhochschule Schweiz, with its laboratory in Web Science, are also participating.