Research On The Fuzzy Clustering Technology Based On Hybrid Character

With the rapid development of the Information Technology and the Database Technology, it is very convenient for people to obtain and store massive information. When dealing with the massive information, people have to face the bottleneck of extracting the useful images rapidly from that information. After ten-year's development, the technology of cluster Analysis, as an important data mining technology, has been applied widely in the fields of artificial intelligence, information control, medical diagnosis, weather forecast, image analysis, and so on.In the previous, the sample was commonly regarded as the point in the feature space and the characteristic of the pattern was determined by the data. Therefore, people usually focus their interests on analyzing the information implied in the data and revealing the basic structure of the set, and they seldom consider the source, the form of characterization and other characters of data sets. In fact, during the process of data mining, people always encounter a problem that the calculation workload grows with the expansion of the sample sets. How to compress and characterize the samples in an effective way so that the process of cluster analysis can be speeded up is the problem worth researching. In the present paper, the author tries to find an effective method of cluster analysis to deal with the mixed characteristics of the samples. The basic research results are as follows:(1)A fuzzy cluster algorithm is supposed to deal with the samples with mixed characteristics, in this paper. In the traditional study, people seldom analyze the relationship among the features of data, and they seldom make the metric analysis on the features of data and the quality of the clustering results. Considering this, the author makes a thorough analysis on the method to realize the cluster consistency and clustering complete and finds out an effective iteration calculation method. Based on this, the author works out a fuzzy cluster algorithm to deal with the samples with mixed characteristics. According to this algorithm, people can compress the big sample set through the analysis of the characteristics of the samples, then make effective cluster analysis on the samples with the same characteristics by the iteration calculation based on cluster consistency and clustering complete, and finally polymerizes the effective characteristics. All in all, the above mentioned fuzzy cluster algorithm comes into being.(2) A reusable software module of fuzzy cluster,which includes the library of function to deal with image, library of function to extract characteristics,module of image data management and fuzzy cluster module base on hybrid characteron ,is supposed to deal with the samples with mixed characteristics in the paper. With this module, people can conveniently increase the operation of extracting various features of data and construct the mixed characteristics collection of clustering consistency and clustering complete. Therefore, this module can be applied to the various cluster analysis on complex and massive data.The fuzzy cluster algorithm put forward in this paper , compared with traditional clustering algorithm ,can be proved to be an effect method by large number of numerical simulation examples and examples of dealing with remote sensing image. Using this method to deal with the remote sensing image, people can select appropriate characteristic from various characteristics and then make effective clustering analysis on remote sensing image.