The Research Of Recognition Technology Based On Fuzzy Rough Sets

In recent years, with the development of science and technology, the targetinformation is more inflation and complex. In the face of such large amout ofinformation and fuzzy information environment, the accurate recognition of target isbecoming more difficult and important. Fuzzy rough sets can effectively discoverimplicit knowledge and classification rules from uncertain and inaccurate information.So the Fuzzy rough method is used in in this paper to realize the target recognition.The main contents of the paper are summarized as follows:Firstly, the basic theory of rough sets and fuzzy rough sets is introduced. Secondly,two core steps called blur and attribute reduction of the identification technology basedon fuzzy rough set are analyzed and studied in-depth.For attributes fuzzification, aiming at the problem of general fuzzification methodbased on fuzzy c-means clustering (FCM) susceptibly influenced by isolated points, animproved FCM fuzzification method is proposed in the paper, and the experimentresults show that this improved method can get a more accurate fuzzification results.Then the FCM fuzzification method is further analyzed, to the problem of FCM needdesignate the sort number and sensitive to initial clustering center, an adaptive FCMfuzzification method is proposed. The experimental results indicate that this methodnot only can automatically adjust the type number of each attribute and get a bestdivide of decision table, but also can improve convergence speed during the divideprocess.For attribute reduction, the widely used reduction algorithm QuickReduct isintroduced and analyzed. Beacause QuickReduct is based on the algebraic view, so it isless intuitive and hard to understand. To introduce information theory to the fuzzyrough sets, the expression of condition entropy is rewritten and a reduction algorithmbased on condition entropy (FRCE) of fuzzy rough sets is proposed.in this paper. Theexperimental results show that FRCE can find a smaller reduction set thanQuickReduct, and it cost a less time.Finally, based on the above research, a SAR recognition system based on fuzzy rough set is designed and it is applied to the SAR image recognition of MSTARdatabase. The recognition experiment shows that the recognition system is feasible,and effective. In particular, when using the method combined by AFCM algorithm andFRCE algorithm, the correct recognition rate is significantly higher than ordinarymethod.