Research on forming line features of basic shapes based on fuzzy cognitive map

Abstract

A method of forming line features of basic shapes based on fuzzy cognitive map is presented to avoid the direct handling of a large number of fragmented line features for shape recognition. The cognitive map, which is constructed by the linking features of the short lines, acts as a reasoning tool to judge whether some of the short lines can be treated as a single long line. The reasoning results of the constructed fuzzy cognitive map would form the meaningful line features, which are component parts of some basic shapes. The concept functions in the fuzzy cognitive map represent the changing range of the composing part of the basic shapes. And the construction of the fuzzy cognitive map network represents the relationships among those composing part of the line features. So, there appears the method of forming the basic line features based on fuzzy cognitive map. By matching reasoning with that model, belief of the combination of short lines into long ones can be obtained. Only when belief is greater than certain threshold, can the lines be regarded as the line features. Experiments show that this method has many advantages, such as simplicity, robustness, small amount of computation, and so on. In addition, this method is comparatively insensitive to rotation, shift and scale of the object. The method we propose in this paper is in a measure the perfection and extension of methods currently used in basic shapes recognition.