Corvus Drones recently collaborated with Track32 to develop an application for measuring the diameter of Belgian Mums to accurately determine their ripeness.
Traditionally, employees manually measure a few representative Belgian Mums using a measuring tape. However, our autonomous drone enables precise measurements of a larger number of flowers in less time, providing growers with better insights into ripeness.
To validate the accuracy of the drone's diameter measurements, we conducted a validation comparing manual diameter measurements with those obtained by our drone. Keep reading for the results!
Our drone takes top-down images of Belgian Mums, allowing the application to detect individual flowers. The application then calculates the diameter of the chrysanthemum in pixels, which is then converted to centimeters. These diameter measurements are averaged and provided to the grower, offering an accurate indication of ripeness.
For validation, Kwekerij Baas in Ens from the Netherlands, was selected as the test location. We chose a total of 20 representative and non-representative Belgian Mums in a mature stage of ripeness. To calculate the actual diameter, each flower was accompanied by a QR code, serving as a known reference for comparing the drone's diameter measurements. The QR code is an easily detectable square reference with accurately known physical dimensions, enabling us to compare the drone's diameter measurement of each individual chrysanthemum with a physical, known reference.
Our drone calculates the diameter by utilizing its height from the ground. This is measured through visual positioning systems on greenhouse poles. This method eliminates the need for QR codes, relying solely on greenhouse pole markers for accurate measurements.
This table shows the actual diameter of each chrysanthemum and the diameter calculated by both our drone application and measuring tape. Overall, the drone's measurements showed accuracy within 2 cm of the actual diameter. Some outliers were observed for flowers that were not perfectly round in shape (marked in the table), which made it difficult to accurately estimate their diameter. On average, the deviation of the drone's measurements from the actual diameter was 0.8 cm, which is considered acceptable. Larger errors up to 3.6 cm were observed for non-spherical Belgian Mums, indicating that there is still room for development in that area.
Improvements and Future Implementation
With this application, growers can measure the ripeness of more plants, faster and more accurately than ever before, surpassing the limitations of manual measurements. The diameter measuring application can be customized for different crops by adding additional training data. For example, lettuce can benefit from this application, as well as any other crops that can be individually detected from a top-down perspective.
The valuable experience we have gained is combined in a practical testing environment at Kwekerij Baas this year. Over the past 6 weeks, an autonomous drone has meticulously monitored an entire bay of chrysanthemums, providing valuable insights into their growth curve. These automated drone measurements will be compared with manual measurements to ensure they are in line with human expectations.
After this validation, the application can be scaled up for larger greenhouse surfaces and potentially even for various crops. This allows for broader utilization of the benefits provided by this technology.
The New Ground Truth in Diameter Monitoring
The validation mentioned in the final paragraph has recently been completed. The drone's measurements have consistently aligned with the growers' expectations, establishing it as the new ground truth for accurate and reliable data. In contrast, human detection has proven to be inconsistent, with different individuals yielding varying results simultaneously and even the same individuals producing different results at different times. The drone's monitoring capabilities have surpassed the limitations of human detection, solidifying its role as the ultimate reference for precise measurements.
We are always open to do a validation check with customers to prove the accuracy of our drone. We’re constantly developing and improving our applications. Do you want us to validate an application as well?