Smart kalanchoe cultivation with cameras, data and AI

Four Dutch pot plant growers have successfully completed a two-year project to develop an AI-based growth model for Kalanchoë, a pot plant widely grown in the Netherlands. The model uses image data captured in the greenhouse to predict the optimal moment for applying growth retardants and harvesting.

The project brought together propagators SV.CO, Vilosa, Slijkerman, and KP Holland, who began collaborating in 2023. They partnered with B-Mex, a Wageningen-based technology company, to develop the Kalanchoë Growth Model. Project management was handled by Florpartners. The initiative originated from a question posed in late 2020: “Where do you see opportunities for precision horticulture in the Kalanchoë greenhouse?”

Over the course of two years, the partners worked toward a functioning demonstration model. The growing facilities of SV.CO at Kreekrug and Slijkerman in Heerhugowaard served as trial locations. At these sites, cameras mounted on spray booms capture images of the crop as they travel across the greenhouse, measuring plant height and maturity. Combined with cultivation plans and climate data, the system generates a prediction of expected height and maturity at the time of delivery. The dashboard advises growers on the correct moment to apply growth retardants and identifies which batches are ready for sale. Machine learning continuously improves the accuracy of the model and its predictions.

Thomas Botman, grower at SV.CO, uses the dashboard daily: “With the dashboard, I have real-time insight into which batches need to be retarded. We no longer have to walk through all the compartments to check.” Roy van der Knaap of KP Holland adds: “Our sales team now knows exactly which cultivars have reached maturity stage 1 at any given moment and can be delivered.”

Testing both the hardware and software under practical conditions required considerable effort. A quarter’s worth of photographic data was collected to properly validate the growth model. Jeroen Boonekamp of B-Mex: “We are pleased to see that the combination of real-time data, AI, and crop models produces reliable predictions of plant height and maturity across different varieties.”

With the technology and model now successfully validated and initial technical issues resolved, the system is being rolled out across the other sites of the participating companies. René van Dop of Vilosa: “With this technology, we can not only improve predictions and reduce our use of growth retardants, but also take a next step towards autonomous cultivation.”

For more information:
Florpartners
[email protected]
https://www.florpartners.nl/