Seafood Fraud Cases
Seafood fraud happens on a vast global scale
Today, more and more cases of mislabeling of fish are being reported around the world. Retailers rely on the certificates they receive with the product. Since it is expensive and time-consuming to check each batch for authenticity, food fraud is very common in this sector.
Not only are customers misled by referring to low-value fish as high-value fish, or by referring to farmed fish as wild fish, but endangered and vulnerable species are also sold which compromises ecosystem stability.
In order to combat seafood fraud and stop giving fraudsters free rein, a quick and inexpensive solution must be available.
Mislabeled fish case
The Xpectrum Pure&Sure device allows the use - and creation - of classification models for species detection. The standard method for this classification is DNA analysis, which is rather expensive, and the results are not immediate. An easy-to-use handheld scanner is very practical since it can be used by any operator in the field.
For this case we focus on frozen fish filets, since these are more difficult to differentiate with the bare eye, but the same method works for fresh fillets. We only focus on species that are interchangeable with one another.
Results
To put a hold on the fish fraud, we analyzed the different types of fish that our customer processes. Which ones look the same when cut as fillets, and which ones can be interchanged with to make more profit. After analysis, we listed the following:
-
Atlantic Cod vs Pacific Cod
-
Atlantic Salmon vs Pacific Salmon
-
Atlantic Haddock vs Atlantic Cod
-
Pacific Cod vs Alaska Pollack
To build the model, the customer had to scan several samples of each grout with the Pure&Sure device, preferably from different batches. Since a scan takes only a few seconds, this could be done fairly quickly. For each spectrum we created, metadata had to be entered into the platform. In this case, that would be the exact fish species. Once enough data was available, the platform used artificial intelligence, data science and machine learning to create the model.
For these cases, we achieve 99% accuracy. We then implemented the solution into our customer's daily operations along with the quality department. Measurements can be quickly taken as input checks by workers or by the quality department itself.
For more information download our free whitepaper on spectrometry for fish products and our own handheld Pure & Sure scanner.