Challenge
Vertical farming is driving a step change in sustainable, low-food-mile agriculture. As farming towers can have growth trays stacked up to 12 meters high, and there can be hundreds of towers, it’s time-consuming to manually identify dying or struggling plants. However, it’s imperative to spot unhealthy crops early so that they can be treated, dying crops removed, yields optimised and resources not wasted.
Solution
We developed a computer vision model to assess crop health by analysing image data from the growth towers. The model flags crops that do not grow as expected. To enable the model, we created a consolidated data environment in Microsoft Azure which brought together numerical data with text and imagery.
Outcome
The growth monitoring model has achieved an accuracy of over 75% and allows the monitoring of crop health to be carried out remotely and automatically. Furthermore, the model is fully scalable and can be used to develop computer vision for any type of crop, allowing the system to be implemented across all growth towers.
