Statistical pattern recognition for cutter positioning in automated fish processing

The authors present an application of computer vision and pattern recognition techniques to the problem of estimating a reference position for a robotic cutter used in the head removal stage of automated fish butchering. The image processing techniques used to locate the necessary features on the body of the fish are discussed. The performance of reference point estimation using multiple regression is presented in comparison with an estimator based on a neural network. It is shown that the system described is capable of improving the butchering efficiency over that of cutters which use fixed average settings, by locating a robotic cutter within 2.5 mm of the optimal cutter location as identified by manually butchering a representative batch of fish.<
>