Computing with words =? New mathematical theory of descriptions. Part 2: Description of patterns (handwritten, face and solid object recognition)

摘要

For pt.1 see ibid., p.657-62 (2001). The problem of computer vision (L. Shapiro,2000) is to automatically characterize the content of digitized images. This problem is complicated by external sources of uncertainty such as the condition of image acquisition. In fact, storage and restoration of images in human brain still obscure. How by imagining data, human stores and restores objects in different dimensions? Nevertheless, a human being uses description in different tasks such as describing a path or an object. For example, one can describe features of a face and our speaker can imagine or recognize the person. The description is made using some words of a given language. Let us imagine the gain one will have in storing, in processing, in transferring, etc., if one arrived to describe an image using some words that a computer can recognize. So, the utilization of fuzzy logic=(computing with words) (L.A. Zadeh, 1997), is necessary. The paper emphasises that description is a robust method for recognition problem solving. In this description, one situates components composing an object in a predefined order. Nevertheless, it is necessary to emphasise that the generation of fuzzy rules here will be different than the classical method (H. J. Zimmerman and P. Zysno, 1985) based on clustering.