«Мы будем держать вас в курсе»
About a year ago, Moba announced a futuristic development program for detection. By investigating the possibilities of the latest technology in visual systems combined with artificial intelligence (referred to as "AI"), Moba aims to set new standards in crack detection. After a year of hard work, our technicians have reported a breakthrough that we consider a game-changer. A method that works on all eggs, no matter what color they are.
Previously, the most effective way of detecting cracks was to perform acoustic tests on the eggshell. It works like tapping on a wine glass: If you hear a bright "pling," the shell is intact — but even the smallest crack will create a "plok" sound if the shell is tested close to this crack. Although this is a very reliable technology, new developments are on the horizon whereby contact with the eggshell is not necessary. That is exactly what can be achieved with a visual crack detector. Another advantage is that cleaning procedures are being significantly simplified, which is especially helpful in markets with very strict hygienic requirements. Numerous companies have researched visual crack detection and some even created commercial systems — but all attempts failed and none of these systems came close to the accuracy of the acoustic crack detector.
A true game-changer is AI. AI networks can be trained to see things that go beyond human recognition capabilities. The challenge is training the AI network with thousands and thousands of sample pictures and teaching it what is right or wrong. And since eggs are natural products, no two cracks are the same. In-depth research has shown that modern AI networks can cope with this kind of variety. This allowed us to overcome a difficult hurdle in our project.
White and brown eggs
The second biggest hurdle is the color of the eggs. Eggs come in all kinds of shades, from pure white to very dark brown. They can be other colors too, such as light-toned, pink, blueish and even green. In theory, this means the training process needs to be repeated for dozens of shell colors, resulting in a never-ending training process and a high-risk profile for every new installation. Luckily, our engineers worked together with top specialists from our strategic partner company Image House Food to develop a method that works on all eggs, regardless of their color. A breakthrough like this wasn't possible before. This technology has been patented and overcomes this hurdle in the project.
Field test at Herbruck's, MI, USA
After such a big discovery in the lab, the most exciting moment was testing it in the field. We were able to perform an extensive field test at Herbruck's in Michigan, USA. It all started with collecting multiple egg images and subsequently training the AI algorithms during the grading process. And it paid off! The detector performed beyond expectations on both white and brown eggs of various housing systems. For a significant time, both the traditional acoustic system and the new visual system worked in parallel, which enabled our team to concentrate on the eggs that divided the two systems' opinions. This resulted in a steep learning curve and, after a few weeks of parallel testing, the grading operation started running solely with the visual crack detection system.
I am excited that our company could work together with Moba in getting the first vision crack detection operational. It was amazing to see that a computer can be trained to detect things the human eye can't see. After Moba trained the artificial intelligence system for several weeks, I trusted the results of the field test unit completely. We were amazed by the accurate results and can't wait for the production of this system to start!
Initial focus on the washed egg market
In many countries, eggs are washed. In other countries, eggs cannot be washed due to regulations. Owing to hygiene requirements and extensive cleaning programs, the first market to apply this technology will be North America, followed by other regions where eggs are being washed. Commercial systems will be available on new Omnia PX machines as of early 2022.
The research program continues
While Moba is starting to design and build commercial units for the washed egg market, the company is simultaneously focusing on its next extensive research program: to investigate the possibility of using similar technology in the non-washed market. Non-washed eggs contain more dirt particles during the crack detection phase and this will bring new insights to scientists.
We will keep you posted on this amazing quest to modernize egg grading using the very latest technology. For more information, please check the product page or contact your local sales representative.