Robo-fishmonger uses 3D images to aim its knife

[Nipuna]

[media]http://www.youtube.com/watch?v=k8miF4psTRk[/media]

Factory robots are now mastering the fine art of filleting fish. Harry Westavik's team at SINTEF Fisheries and Agriculture in Trondheim, Norway, is developing a system that can fillet farmed fish with the help of a laser that captures 3D colour images of the animal.

Algorithms analyse the images to work out the best place to cut, before a robot arm takes over. The system can be trained to distinguish between different species, allowing it to sort fish by type. It can also determine weight by estimating volume from images, then applying the typical density of a fish.

While, the goal of the project is to replicate the performance of a human operator, experienced workers are still more effective at the moment, says Westavik. "The biggest technical challenges are being able to detect internal and external fish quality and structure and finding tools that can handle the fragile texture of a fish," he says.

Because of the variability of individual fish, new techniques must be tried out on real specimens. The team is working on a fully automated production line that can handle all aspects of salmon trimming, to be trialled in 2016.