High-speed model-based weight sensing of complex objects with application in industrial processes

Advancement in food processing requires development of innovative sensor technologies that are flexible, fast, and accurate. A new approach has been developed for achieving optimal packaging of product portions so that the deviation of the portion weight from a specific target weight is minimized. The new portioning technology depends primarily on fast and accurate sensing of complex properties of the processed natural object. Conventional sensors would not meet the speed and cost constraints of on-line processing. A new model-based approach for fast and cost effective on-line sensing of a complex property such as the weight distribution of a natural object is presented in this paper. The approach involves off-line development of a model-base that contains non-dimensional models of the target property and a set of correlated simple features that can be measured easily at high speed and high accuracy. The on-line sensing process involves measurement of these simple features for each processed object and matching them to the appropriate property model in the model-base. The approach is applied here in an innovative can-filling process of fish to estimate the weight-rate function of the processed fish. Real data of fish are used to build the non-dimensional models of fish. Results presented in this paper indicate that the sensing approach is suitable for the fish processing industry with respect to the needed accuracy and speed.