Genetic algorithm for optimizing fuzzy image pattern matching

Fuzzy image pattern matching recognizes and localizes objects in images using methods of fuzzy logic. The methods use fuzzy segmentation, templates and operators that bring a heap of alternatives into the reasoning process. Fuzzy pattern matching takes into account uncertainties and inaccuracies of the input images as well as the template images and incorporates them in the matching process. In this paper, a method for fuzzy pattern matching and a method to optimize the matching scheme using a genetic algorithm are proposed. Optimization aims at finding reliable features from a set of calibration images by simultaneously optimizing the segmentation function and the template. Optimization proved to result in good generalization of the matcher for the unseen test images and better performance than the common correlation pattern matching method.