Picking cucumbers will soon be easier than ever thanks to a Dutch-made robot. Crux Agribotics' fully automatic harvesting robot also fishes out the cucumbers hidden deep among the leaves.
The harvesting robot is best described as a trolley with spinning cameras and a flexible gripper. The gripper will only do its job once it has looked through the leaves and the robot has a clear picture of where the cucumbers are hanging. The cameras in the robot take quick and comprehensive 3D images of the crop based on these findings, and specially developed software puts all those findings together and draws a conclusion about where exactly the cucumbers are trembling in the foliage and what the shapes and sizes are. "The coordinates are then transmitted to the gripper, which clamps the stalk and simultaneously cuts through the top," director Richard Vaille tells The Engineer. ''That way, we grab about 96 per cent of ripe cucumbers. For the remaining, difficult cases, a limited number of employees walk along." According to The Engineer, Vaille did not want to say much else about the grabbing and cutting mechanism.
Self-learning algorithm
The harvesting robot is not at that 96 per cent level overnight: it learns by trial and error through a self-learning algorithm. Using imagery, this algorithm is told which cucumbers are ripe for picking. The learning process is partly the grower's responsibility. "For example, the robot sends him a picture with a tricky case and asks: what do we do with this?"
Similarly, the robot learns to detect diseased cucumbers. "That saves quite a lot of pesticides and crop loss. People often catch it too late, so many cucumbers are quickly affected by a virus and have to be sprayed."
The algorithm is additionally capable of making predictions. It then tells the sharer, for example, that a cucumber will be ripe for picking in three days.
Organic products
"The difficult thing about what we do at Crux is that we deal with organic products," Vaille states in a company video. "Think of cucumbers, which have a different shape every time. We cannot integrate that into our software by reading the drawing of the product. So we have to look for unique shapes every time. That means we are going to take the next step in technology, for example machine learning. The more cucumbers we see on our camera images, the better, smarter and more accurate our algorithms and software become."
The harvesting robot is just about ready to go, according to its director, all showstoppers - anything that could jeopardise progress - have been eliminated. The company is therefore also setting its sights on a new packaging robot. "That would mean virtually no more hands involved," he says.
The prototype of the harvesting robot has already been tested at several growers, and Crux Agribotics is now turning it into an industrial product after which the complete platform from harvesting to packaging should hit the market in 2019.
By: Kelly Bakker
Source: Crux Agribotics/De Ingenieur
