Abstract:

To the issues of target recognition (TR), a technique for TR based on adaptive neurointuitionistic fuzzy inference system (ANIFIS) is proposed with intuitionistic fuzzy inference—neural nets theory introduced into the area of information fusion. First, after analyzing the properties and vulnerabilities of the existing TR methods, ANIFIS is proposed. Moreover, because the logical system can be mapped a fuzzy multilayer feedforward nets system, a model for TR on ANIFIS with TakagiSugeno type is established. Then, the attribute functions, i.e., membership and nonmembership functions, and the inference rules of the system variables are devised with computational relations between layers of input and output and a synthesized computational expression. Subsequently, a learning algorithm of neural net is devised to train net and modify rules. Finally, the output results and recognition precision based on two techniques, including intuitionistic fuzzy inference and ANIFIS, are analyzed and compared by providing TR instances with 20 typical targets. The simulated results show that it is a more practical and valid technique on decisionmaking fusion which can improve recognition precision and training speed.