The Research On Fuzzy Clustering Analysis Technology And Its Application

The fuzzy clustering analysis technology is an important research direction in the intelligence information processing and an unsupervised fuzzy pattern Recognition method too, which is appling to many fields. It researches clustering question which uses the fuzzy mathematics algorithm. The thesis thoroughly studies fuzzy c-means clustering algorithm principle and applies it to the image and the made dot sample set. The thesis finishes the following work:1. The thesis studies the traditional clustering algorithms, presents the basic idea and process of the algorithm of the hard c-means clustering and the fuzzy c-means clustering. It also presents the fuzzy c-means clustering algorithm basing on improving the condition.2. The thesis studies the fuzzy c-means clustering parameters, and detailedly discusses the clustering center initial method question, ascertaining fuzzy power index m, chooseing the distance space and objection function astringency question. The thesis proposes two methods of ascertaining fuzzy clustering number c, and applies the method to clustering in image and the made dot sample set.3. Five kinds of binarization algorithms are presented, which includes maximum square, maximum between-class cross-entropy, minimum cross-entropy, maximum between-class fuzzy divergence and minimum fuzzy divergence. It compares the binarization algorithms with the fuzzy c-means clustering algorithm.4. Basing on sugeno's g_λ fuzzy measure clustering algorithm, the thesis proposes a weighting fuzzy clustering method and gives out the weighting value's calculation method. The two kinds of weighting fuzzy clustering method are applied to image segmentation by utilizing one-dimensional gray feature of image and one-dimensional gray histogram which is weighting value, two-dimensional gray feature of image and two-dimensional gray histogram which is weighting value. The thesis also presents the result comparison.5. Aiming at gray image, the thesis compares the result of the fuzzy c-means clustering algorithm and the self-organizing feature mapping neural network method.