Flanders Make develops drones for logistics and agriculture
Drones can take over time-consuming tasks that people today still often perform manually, completely autonomously and with the help of artificial intelligence. For example, a drone can be used to count objects from the air using cameras and sensors. Flanders Make has developed drones that automatically take stock in a warehouse and an algorithm that can count the number of flowers and fruits on a piece of agricultural land to make a harvest forecast.
Thanks to new technological developments such as smart drones, companies can make various processes safer, faster and more efficient. After the successful inauguration of the Flemish Drone Federation EUKA at Flanders Make a year ago, Flanders Make now shows two applications with flying robots. For farmers, for instance, it is crucial to make a good harvest forecast. This is very interesting to make concrete agreements with customers in advance and to estimate how many resources are needed for the harvest work. Thanks to artificial intelligence, drones can now count the number of strawberries from the air, helping crop growers to predict their harvest more accurately.
Counting strawberries from the sky
"For this case, we collected aerial images of through-bearing strawberry crops," says Rob Heylen, researcher at Flanders Make. "Our algorithm analyses the images and counts the number of fruits and/or flowers that are visible. By also doing a manual count during our research, we were able to determine an error margin between the number of digitally recognised flowers and the actual number of flowers. Because we now know this margin of error, we are able to determine the actual number of flowers very accurately based on quickly collected drone images."
This application was created in cooperation with the Experimental Centre for Fruits Growing. They were already doing manual counting of strawberry flowers to predict the harvest yield of flowering strawberries over a period of 3 weeks. Thanks to the drone, this can now be done much faster. An important factor for the successful acceptance of new drone applications in the business world is the use of artificial intelligence. This technique enables drones to carry out tasks completely independently or to analyse collected image material at lightning speed.
The same kind of technology can also be used for inventory management. Flanders Make, together with IMEC, has developed a drone that races through the racks of a warehouse at an insane speed of no less than 4 m/s to count the stock. In this way, an entire aisle of a large warehouse can be inventoried in just a few minutes.
"We have developed a flying robot in our own drone lab that can scan the stock in a warehouse completely autonomously on two levels, says Jeroen Zegers, also a researcher at Flanders Make. "On the one hand, we use digital cameras that recognise QR codes on the shelves and the pallets in order to make an inventory of the number of pallets in stock. We soon noticed that there was a demand from the market to be able to count the number of individual boxes as well."
Therefore, the researchers have developed a second application that, with the help of LiDAR technology, can also count the number of boxes on a partially unpacked pallet. A LiDAR sensor measures the distance between the boxes and the drone. By repeatedly removing the boxes from front to back and top to bottom, the drone can quickly calculate the number of boxes remaining. A job that is normally very labour-intensive, but that the drone now carries out in the blink of an eye.
Counting other objects using drones
The drone applications that Flanders Make has developed for agriculture and logistics are now being applied in practice but are also applicable in other sectors, on other types of objects, provided the algorithm is retrained. So companies that have business cases where it would be interesting to count objects automatically from the air can always contact the research centre.