Machine Vision Based Fish Cutting Point Prediction for Target Weight

Abstract

Food processing companies pursue the distribution of ingredients that were packaged according to a certain weight. Particularly, foods like fish are highly demanded and supplied. However, despite the high quantity of fish to be supplied, most seafood processing companies have yet to install automation equipment. Such absence of automation equipment for seafood processing incurs a considerable cost regarding labor force, economy, and time. Moreover, workers responsible for fish processing are exposed to risks because fish processing tasks require the use of dangerous tools, such as power saws or knives. To solve these problems observed in the fish processing field, this study proposed a fish cutting point prediction method based on AI machine vision and target weight. The proposed method performs three-dimensional (3D) modeling of a fish’s form based on image processing techniques and partitioned random sample consensus (RANSAC) and extracts 3D feature information. Then, it generates a neural network model for predicting fish cutting points according to the target weight by performing machine learning of the extracted 3D feature information and measured weight information. This study allows for the direct cutting of fish based on cutting points predicted by the proposed method. Subsequently, we compared the measured weight of the cut pieces with the target weight. The comparison result verified that the proposed method showed a mean error rate of approximately 3%.