In Hellevoetsluis, the Netherlands, Eqraft’s new electronic onion sorting machine was presented last week. This line could subsequently be seen in action at the company of seed improver Jim Hoogzand, who has the first optical grader in use.
The launch of this significantly renewed and improved machine coincides with the rebranding and merger of ERC Machinery, Qreenno and Propak into the new company Eqraft. In party centre Ri-Jo in Hellevoetsluis, Rutger Keurhorst and Gerard van der Knijff from Eqraft opened the afternoon with a short explanation about that decision and about the developments within optical sorting.
Bas Pomstra, head of R&D within Eqraft, then spoke to the development team about which improvements have been made to make the machine less maintenance-sensitive and easier to use. In addition, a number of new techniques have been added in the field of vision, so that the machine generates better images for the external quality sorting.
Eqraft’s Optical Grader sorts onions internally and externally at high speed, based on colour, shape, weight and both internal and external quality. During this process, the onions first travel over an elevator belt, after which they are singled out and placed in cups one by one. Modesta’s Blower Beam removes loose skins and possible dirt from the onion. A conscious decision has been made to continue with the concept of sorting onions in a cup, collecting data from an unmoving product and preventing wear and tear and unusable images because of, for example, rollers. Furthermore, a great number of improvements have been made at detail-level based on user experiences of existing Longobardi customers. Precisely because of that coupling with the years of experience in the onion industry within Eqraft, the company can now present a sturdy Dutch machine it has every right to be proud of. Each cup is individually driven, which is why the machine knows the parameters of the onion per cup. At the end of the entire process, this selection is placed in crates from six different exits from the machine, and each crate will be filled with the desired type of onions.
Rolf van Beek from Q.E.F., which has been working closely with Eqraft for decades already on the development and improving of image analysis on other machines, such as for example, the fully automatic onion peeler, explained which additions in data analysis make this new machine so special.
Thanks to a new method of image analysis besides the existing software, which is already quite good, a computer could be trained, as it were, to not just recognise a certain characteristic, but to also understand it. This is the same sort of technique that is used in, for example, a Tesla, which can learn to recognise and judge new situations by loading thousands of traffic situations into the system. Thanks to this technique, the Optical Grader can sort onions by decay, size, shape and more.
Rolf: “Every company looks at an onion differently, and wants to sort it based on different characteristics. This machine can be ‘trained’ to do what the customer wants, and learns to understand the product in that manner. However, this additional layer of software is not yet functioning at the level we would like it to be.” Rolf gives a rough estimate of approximately 90 per cent. Much data is needed to go through this learning process, and he then encourages the room to participate and think along and supply products. Within several months, enormous steps could then be made, such as have been made in the software to recognise handwriting which they developed, and which now is 99.9 per cent accurate.
After the presentations, the audience, all professionals from the fresh produce sector, had some critical questions: does the machine work with red onions, it is profitable in the onion branch, and what is its accuracy percentage? The men from Eqraft gave comprehensive answers to these questions. It was good to experience how the technicians, who tend to be somewhat more careful, seek input and cooperation from the sector to go further.
Bas: “The machine has to be further developed and ‘trained,’ but the fact is that it is already more consistent and accurate than human counterparts.” This was confirmed from the audience by Kees Hoogzand, who concluded that the machine is currently already sorting as well as humans, and that this will only improve in coming months. “The eventual goal is that the reading rooms will be empty, and this machine is capable of making that happen.”
HOZA, Jim Hoogzand’s breeding station, has already had the first Optical Grader in use for some months. After the presentation, this could be personally experienced by the audience: the working machine could be looked at, and there was room for questions. After that it was time for refreshments. The conversation continued over pizza, beer and pop. Rutger: “We originally planned to do the presentation and the tour around the new Optical Grader for about five or ten people, but interest was much higher than that. The room was filled with 60 people, and of those invited, about 80 per cent turned up; this indicates that these new developments are alive in the sector.”
What will the next step be? Eqraft has started a course of participation, in which the Optical Grader will be further developed. It truly has to become a machine from processors, for processors. As regards image analysis, enormous strides have been made, with calculation capacities which were not yet possible two years ago. The challenge now is to arrive at an independent classification of onions, so that each cultivator can put together his own recipe. Additionally, the participants will get together about four times a year. Feedback and discussions will be encouraged during those meetings: that is how the machine will be further improved. During the next meeting in April, the first results of the learning course, which is in full swing behind the scenes, will be shared, and the difference with current analysis techniques will be shown.
