Digitally grow peppers and detect lameness

Virtual Plant Twin

GreenTechLab is located within Fontys Venlo and deals with technical questions within agro, food and logistics. Digital twinning offers a unique challenge in this; where in an industrial process, IoT solutions can collect data and link it back to reality, this is not possible with a living organism. Applications of digital twins in agro and food therefore work differently, but can mean a lot in process optimisation, data collection and signalling. Roosen and his team started two pilot projects to explore this; lameness detection in animals and digital cultivation of peppers.

Animal welfare with digital twins

Livestock farming often involves keeping large numbers of animals, making timely detection of health problems a challenge. Walking behaviour is a key indicator of possible lameness, which can have far-reaching consequences. Spotting this in time can make all the difference, explains Roosen: "A digital twin of a cow can help analyse video images quickly and effectively. What we have done is created a system that can digitally reproduce a cow's images. In other words, a virtual environment that creates a digital twin of the cow. We trained that system to recognise abnormal walking behaviour in cows. The technique you use for this is called recurrent neural networks, which you can now use live to identify unhealthy walking behaviour. This benefits animal welfare and has many applications in animal husbandry, including for other conditions."

Digitally growing peppers

A similar application has been explored for fruit and vegetable growing, developing a digital twin of pepper plants. This is challenging because this is an AI application that has to learn to recognise peppers: "We recognise peppers instinctively, based on experience and knowledge of the farmer. A system has to learn that first and that takes a lot of time, but it also offers a lot of opportunities. That needs perhaps thousands of iterations, also to enable classification. Indeed, what we get here is a translation of what is in the greenhouse into understandable and analysable data. How many peppers are growing in the greenhouse and how well are they growing? What is their density? Where are the focus areas? That is the digitisation that is missing in cultivation." For the grower, this has many advantages; more insight and predictability in the harvest, better planning and preventive action to keep the plants healthy. Healthier business processes, in other words.

Accelerating digitisation and the 'digital greenhouse'

The technique of image recognition (convolutional neural networks) used in the pepper greenhouse has been around for some time. The innovation is in the application, where the contents of a greenhouse are converted into a dataset that offers many possibilities. Growers can go data-driven on this. Digital twinning can also have a signalling function in disease management; a disease in plants starts small, but often grows exponentially. A digital twin accelerates digitalisation opportunities and makes a digital greenhouse possible, argues Roosen: "In theory, you can reproduce the whole greenhouse as a digital twin and control it digitally. From humidification and air circulation to the day and night cycle. The big challenge here is trust in the system and acceptance of digital twins and AI applications. For an entrepreneur, technology is like a 'black box'; the process cannot be explained. With Virtual Plant Twin, we are taking a step forward to prove the validity of technology."

Digital Twin Academy

Virtual Plant Twin is a pilot project within Digital Twin Academy. Within this larger project, digital twins are made accessible to SMEs and applications of the technology are explored in a number of projects. This specific project is carried out by the knowledge centre GreenTechLab, part of Fontys Hogeschool Techniek & Logistiek, with support from other institutes, including Fontys Hogeschool ICT and Fontys Hogeschool Engineering. Interested parties can find more information and contact us at fontys.nl/dta.

The Digital Twin Academy is an Interreg Euregio Meuse-Rhine (EMR) project. The programme is financed by the “Europees Fonds voor Regionale Ontwikkeling” and is a collaboration of various public authorities, research and education partners and business partners.