Intelligent fusion of sensor data for product quality assessment in a fish cutting machine

This paper presents two intelligent sensor fusion techniques, which have been implemented in an automated machine for mechanical processing of salmon, to determine the level of product quality (i.e., the quality of processed fish). An automated fish cutting machine with advanced sensor technology is employed in the present work. The fish cutting process is complex, and ill-defined, and quality assessment methods are subjective. Two knowledge-based fuzzy fusion methods based on: a) regular Mamdani dot-max composition, b) the degree of certainty are implemented to achieve improved results. The data available from disparate sensors like CCD cameras, optical encoders and ultrasonic displacement sensor of the machine are fused using the two methods. An illustrative example for a good and a bad cut is presented. The results indicate that the two methods are equally effective, but method (a), which is more sophisticated, has a slight advantage in performance over the other, at the expense of added complexity.