Quick automatic sorting in agriculture

Manually removing damaged or diseased potatoes from the harvest is a labour-intensive and complex job. Martijn, Lars and Stefan Flikweert of Flikweert Vision Technologies decided to develop a potato sorting machine that automatically sorts potatoes based on vision and artificial intelligence. To achieve the required speed and flexibility, Festo supported in choosing the right cylinders, controls and bus system. The end result is a machine that can process up to ten tons of potatoes per hour.

In Ouwerkerk, Zeeland, less than three years ago, cousins Martijn and Lars Flikweert started the company Flikweert Vision Technologies with a clear goal: To develop practical solutions that will make the agricultural sector more sustainable and automated. A logical focus having parents who themselves have an agricultural business and educational backgrounds that fit into this theme: Martijn, for example, studied Agrotechnology & Management and Lars completed a Biobased Engineering (agrotechnology) University course. Not much later, Martijn's brother and mechanical engineer Stefan joined the company.


Based on their background, in recent years they have been working on a machine that can automatically sort potatoes. In most agricultural companies, this is done manually to make the batch of potatoes as clean as possible and give the highest value. During sorting, clods of earth, stones and diseased or damaged potatoes are removed. A labour-intensive and therefore expensive affair that is also far from perfect and also allows potatoes with less visible 'disorders' through. Especially when the grader in question has less knowledge or expertise.

Automation, according to the Flikweert men, was achievable by using vision technology combined with AI. This involves using cameras to view the potatoes on all sides in a short period of time and generate a large number of images that are reviewed by AI. With this, potatoes with zip needles (a type of worm that makes small holes in the potato), scab, damage, shape defects, growth cracks and greenery can be distinguished but also previously mentioned mud clods, stones and other types of contamination. Martijn Flikweert: “By first 'measuring' enough potatoes in advance and collecting data, you can make the AI system so intelligent that it reliably recognises these abnormalities. And the more data you keep collecting after that, the better the system becomes. Eventually, this also allows you to clean up very 'dirty' batches of potatoes in such a way that a practically worthless batch can be upgraded to a batch with value. Practice now shows that lots up to 50% 'tare' are salvageable.”

From metal to plastic

With the knowledge and skills of these three men, hard work was done on a prototype. Martijn and Lars developed the concept involving vision and AI while Stefan got to work on the construction of the machine. Martijn: "The basic idea was to bring potatoes into the machine via a roller conveyor. This roller conveyor separates them sufficiently before they come under the vision system where they are then rotated in such a way that the cameras 'see' each potato five times. These images are combined by the software and evaluated for quality. Based on the customer requirements, this software eventually draws a conclusion with two possible outcomes: approve or reject. In the latter case, the potato - or contamination - is removed from the stream using a pneumatically driven 'spoon'.”

Martijn: "Once the machine concept was on paper, it was time to build a prototype. For the drive of the spoons, among other things, we ended up at a technical wholesaler. This is where we first came into contact with Festo components and ultimately with the company itself. An excellent course of action since Festo had just that extra bit of knowledge and experience available that we did not yet have.’

Speed and flexibility

One of the main bottlenecks to solve in the design is the speed and flexibility with which the machine must respond to the conclusion of the vision software. Martijn: "The potatoes are assessed in real time after which a scoop may have to be controlled immediately to remove the potato from the stream and push it onto an underflow conveyor. This spoon is operated by a pneumatic cylinder that must therefore also receive the information quickly and respond to it quickly. We had already looked at using a valve island because we end up working with 42 spoons controlled by 42 cylinders, but exactly how to do this was not clear.”

In cooperation with Festo, a solution was realised in which CPX-E with EtherCAT is the overarching controller that works together with the vision PC. When the vision software determines that a potato should be removed from the flow, this conclusion is translated into the appropriate command towards the PLC. This controls two VTUG islands via the integrated CPX-AP-I decentralised I/O system, which in turn are linked to 42 DSNU round cylinders.

Decentralised I/O system

The CPX-AP-I is integral to this solution. Namely, this decentralised I/O system is able to directly connect the I/O of the valve islands in the field without the need to provide the individual components with a fieldbus node. Martijn: "With this specific solution, we can achieve the desired speed and flexibility required in this sorting machine. But the solution is also easy to assemble and offers the ability to scale up or down almost indefinitely as needed; the control architecture itself is fully preserved. Moreover, it saves mounting additional components that you would need if the valve islands did connect via an IO-Link master. In this solution, we deal with one power cable per valve island and one bus cable to which we can connect cylinders. The quality of the compressed air is monitored and maintained via the air care unit.'

Doing it yourself

Martijn, Lars and Stefan Flikweert are particularly pleased with the cooperation with Festo. Martijn: “They have helped us enormously in making the right choices so that we now have a machine that meets our requirements, in which the control of the scoops works optimally with the vision system and with which we have all the possibilities to vary in the future generation of machines. We learned a lot from Festo and it was also very pleasant that we had direct contact with the right departments. No complicated constructions via account managers, just call, discuss and move on. In addition, the sales consultants at Festo are also extremely well educated technically, making the entire company a good sparring partner. We now really have the idea that we can continue independently...with Festo products, of course. The first ten machines are now with their new owners and a second series is now in the works.

March 2024