Similarity Measure And Pattern Recognition On Geometry Outline Of Multidimensional Data Object

The similarity of sample and its measurement method research become basic work of pattern recognition research, the core of the similarity analysis problem is the simila-rity measure.Analysis from geometrical characteristics of the multiple scale data (sample geom-etry), studys on similarity problems which bases on the similarity algorithm of multidi-mensional data object, and the main work is as follows:Firstly,the concept of outline of sample geometry is given, and resolves the outline of sample geometry associated with sample vector from essential; defines derived simplex based on regular simplex.Secondly, contour similarity algorithm is introduced, and the simulation method of contour similarity mapping are studied.Further mo re, us ing small sample to generat large sample which obey same distribution as test samples, through comparing results of discriminant which is about contour similarity with other similarity measure methods, gets that criterion of contour similarity measure is stable.Finally,the model of clustering and discriminant are established which base on con-tour similarity algorithm, and compares with other methods. Through examples to disc-uss the dimension of sensitivity which is about order of magnitude and dimensional,and variation value.