The Improved Fuzzy-Clustering Approach To Fuzzy Modeling Method And Its Application Research

Fuzzy modeling system is an important branch of modern control theory, which has been widely used in various fields and continuously made new achievements. Fuzzy modeling can identify structure and parameters of the fuzzy model from the input and output datas so as to describe the non-linear system. Image segmentation is based on color, texture, gray and movement, and other characteristics. The purpose of image segmentation is to divide the visual images into meaningful and properties simila region, the use of the image segmentation is involved in almost all the areas.First of all, the article has studied the development process of fuzzy modeling, and made a discussion and research on the hierarchical fuzzy clustering algorithm which is used in fuzzy modeling and image segmentation.Considering the problem that using the nearest neighbor fuzzy clustering algorithm can overcome the clustering center of arbitrary initialization. the article takes the chaotic nonlinear dynamic systems model of T-S fuzzy model as the object of study, improves the fuzzy clustering algorithm through the introduction of the local division of correlation factor to smooth vector samples of each group enter through a series of steps to optimize the T-S structure so that can descript the dynamic characteristics of systems. This method improves the robustness, modeling accuracy and speed of the system modeling.And then because of low speed and low accuracy for fuzzy modeling and image segmentation by using the general fuzzy clustering algorithm, the article introduces the limited space theory to achieve de-noising, as well as for the slower low-speed image noise and poor quality of the division through smoothing input samples near the center point of the input sample, improves the speed and accuracy of the modeling of chaotic dynamic system and image noise removing. Finally, the article improves the hierarchical fuzzy clustering further through the introduction of Gaussian kernel, so as to smooth input sample value of the images and nonlinear system and reduce noise caused by interference. This method establishes a relationship between the image pixels and its surrounding pixels. This method brings high accuracy and speed of the chaotic dynamic systems modeling and image segmentation.