Dynamic fuzzy-reasoning-based system and its applications to food quality assessment

This thesis extends the conventional correlation of instrumental measurements and sensory attributes of a food product to a generalized correlation problem in the framework of fuzzy set theory. To solve the generalized correlation problem, this thesis develops a computer-based system called DFS-based Fuzzy Classifier (DFSC). It can accept a mixed pattern that may contain both instrumental and sensory measurements of a food sample, and give fuzzy evaluation of the sample in terms of the overall quality and the sensory attributes.; The core of the DFSC is Dynamic Fuzzy-reasoning-based System (DFS) that is the new approach developed in this study. The DFS simulates and extends the local heuristic interpolation with the help of fuzzy set theory and vector space theory. Compared with other present model-free approaches, the DFS is characterized by its straightforwardness, flexibility and effectiveness. Little training and knowledge acquisition are needed for the application of the DFS to a particular problem. Experiments with a number of real-world data in system simulation and classification problems validated the DFS and its advantages over conventional fuzzy systems.; The DFSC was applied to the quality assessment of roasted peanuts. The results from the four case studies showed that DFSCs solved the generalized correlation problem in the quality assessment of roasted peanuts. Different input formats were tested, and the performances of DFSCs were comparable to the panel's evaluation. Based on the experiments, we suggest to use color measurements as the input to the DFSC for the quality assessment of roasted peanuts.