The Australian strawberry industry has been hit by ‘sabotage’ this year: recent weeks consumers have over a 100 times found large needles stuck into the berries. To ensure the confidence in Australian berries, the producers have therefore begun to control strawberries for export with X-ray. X-ray inspection of food
is an old and well-established technology, but today you have to develop unique solutions for each product to be able to test for different errors. For example, one solution is used to find out if a potato is hollow, while an other solution is used to find out if cold cuts are contaminated with metal shavings. Therefore, X-ray inspection today is preferably used on expensive products with known types of errors – or to sabotaged strawberries.
But people behind a new Danish project, are working on a solution, where food will be able to be scanned with cheaper all-round inspection machines in the future. The Danish project will develop a new dynamic X-ray technology that can continuously vary the X-ray energy of the device and choose the right camera technology. By adding artificial intelligence, the goal is to eliminate the need to develop unique software solutions for each product, and instead enable the system to automatically distinguish good products from contaminated.
The system chooses by itself!
“We develop artificial intelligence algorithms that can choose the optimum x-ray power and right camera with the right resolution. This means that we can control many different types of food without changing the inspection system”, says Brian Vinter, professor at Niels Bohr Institutet. Aarhus-based Magnatek is responsible for the development of a new type X-ray source, while QTechnology from Copenhagen is developing cameras for the project. Newtec Engineering in Odense is responsible for system integration, and Niels Bohr Institute is heading the software development.
The Danish Technological Institute will validate the final solution, and the project has a total budget of 17 million DKK. “One of the big challenges is to get X-ray sources and cameras to communicate with our algorithm at very high speeds. For example, we work on detecting hollow potatoes, which needs to inspect 22 tonnes per hour. Our plan is to try to make a hardware solution, so image recognition takes place directly in FPGA chips, so we do not have to have a large server standing between production lines, “says Brian Vinter. When an X-ray inspection system can handle many types of food, the price will also be lower. Therefore, Brian Winter hopes that on the long-term, inspection systems like this also can find their way to supermarkets, to make extra quality checks before the goods arrive on the shelves.
If you want to read about a company that are applying AI on food production lines, then read this interview with Rufus from Sensomind
Original article from https://ing.dk/