The only good fish is fresh fish
An optical AI nose sniffs out salmon and co
Hyperspectral analysis, an optical analysis method, and a bit of artificial intelligence (AI) are making it possible to automatically ensure the quality of fish. It’s an interesting approach for the fish-processing industry.
How do you tell if fish is fresh? It’s really very easy. The fish should have a neutral smell and firm, elastic flesh, the eyes should not bulge, and the skin should be damp and shiny. With their senses and a bit of experience, humans can easily distinguish between a fresh fish and a rotten one. It’s harder for machines, but even they can become fish experts thanks to an optical analysis method and a bit of artificial intelligence (AI).
Customer reference: RaisioAqua, Finland
Digital transformation with Siemens Digital Enterprise Services
Trust, expertise, and a holistic digitalization approach: With these “ingredients,” Finnish fish feed producer RaisioAqua is embarking on a digital transformation journey with Siemens Digital Enterprise Services at its side. Which role do Consulting, Implementation and Optimization play in this? What is the visualization of production data in MindSphere and the Internet of Things (IoT) all about? And how does that impact product quality? Find out what’s possible when visions meet technology!
EU project automates the processing of fish
In large cold-storage warehouses, where the deliveries made by the fishermen for retail stores are checked, resorted, and distributed to supermarkets, fish is usually handled as follows: The ice-covered seafood is delivered in large plastic crates, taken out of the ice manually by the warehouse staff, weighed, and repacked in the retailer’s crates. This laborious manual work will now be automated.
“Oceans cover more than 70 percent of the earth,” says Torill Østingsen of Siemens Industries in Norway. “Although only around two to three percent of the global food supply comes from the sea, our oceans have been drastically overfished. And yet the world’s population is growing, and demand for fish, a healthful food, is increasing. We must have (sustainable) aquacultures if we are to protect the oceans while also meeting demand. And automated processes are needed if we are to ensure that only the best quality fish will reach consumers in the future.”
EIT-Manufacturing , a knowledge and innovation community within the European Institute of Innovation and Technology (EIT) founded by the EU, is therefore funding a project in which multiple industrial enterprises and research institutions are jointly developing a solution to this problem. “In the future, robotic fingers will grip and repackage the slippery, sensitive fish,” explains Sven Dudeck from the Siemens Technology Research Unit. “The robotics manufacturer KUKA and the French research company CEA are looking into ways to automate the fish handling procedures. We at Siemens ensure that only top quality fish is supplied to the retailers. In other words, the fish must always be fresh.”
A look at fish skin
Siemens researchers are taking a special look at the fish skin. As soon as the fish is killed, its skin slowly begins to change. The smooth, shimmering skin moistened by the water is altered and loses its elasticity. “During this process, the molecules on the surface of the skin change, and a thin layer of mucous forms on the skin,” explains Alexander Gigler, also from Siemens Technology. “The proteins break down, and new chemical substances form, which are easy for the human nose to detect. We make use of this biochemical aging process. We use hyperspectral analysis to analyze the surface of the fish’s skin. This means that we irradiate the skin with electromagnetic waves whose wavelengths are between 900 nm and 1,700 nm in the infrared spectrum, which means between visible light and thermal radiation. We observe how much of the different wavelengths is absorbed by the surface molecules. Different molecules show typical absorption patterns. We can then infer which molecules are present in what quantity on the fish skin. We’ve used training examples, that is, fresh and not-so-fresh fish, to train an AI algorithm. If the AI then evaluates the absorption pattern of a specific fish, it provides a very accurate statement on whether the fish is still fresh-caught, still edible, or is already of substandard quality.”
2021 launch in Azambuja
This technology is expected to be used for the first time by project partner Sonae MC, the leading grocery retailer in Portugal, and the largest for fresh fish. The Portuguese eat quite a lot of seafood on average, and Sonae sells approximately 53 metric tons of fish each day in its more than 570 stores. The products are first sorted and repackaged in a cold-storage facility in Azambuja (a city in Portugal). A pilot system is currently being built here for the hyperspectral analysis, which will be placed into operation as early as 2021.
Aenne Barnard, April 2021
Subscribe to our Newsletter
Stay up to date at all times: everything you need to know about electrification, automation, and digitalization.