A improved k-means clustering algorithm combined with the genetic algorithm

The cluster approach is one of the basic methods to pattern classification and system modeling. The clustering target is that according to some rules, divide the sample data set in the sample space into some subsets indicating different patterns or system behavior[1]. In the course of establishing the video image indexing, every step from building index according to the basic visual characters of extracted images, to forming category index trough extracting related programs, uses the clustering thought. So, how to choose a suitable effective clustering algorithm will directly affect the efficiency to establish the video image indexing and the performance of the whole management system. The k-means clustering algorithm is a relatively good one.