关键词: AdaBoost.M2, 神经模糊系统（NFS）, 植物识别, 支持向量机, 隶属度

Abstract: In order to improve the recognition rate of Neural-Fuzzy System(NFS) towards similar samples, AdaBoost.M2 was proposed. Firstly, the traditional NFS output layer was removed to obtain a new NFS because of the requirements of Adaboost.M2 to weak classifiers. Secondly, with the combination of AdaBoost.M2 and new NFS, a new AdaBoost.M2 Neural-Fuzzy System(AdaBoost.M2-NFS) model was obtained. It is suggested in this experiment that the recognition rate of the new model is 3.33% higher than that of the single NFS. In addtion, the recognition rate of new model is 1.11% and 3.33% higher than that of linear Support Vector Machine (SVM) and Softmax respectively. Based on sensitivity and specificity analysis, the non-linear data can get better result of classification than the linear data. The results indicate that the AdaBoost.M2-NFS-based model not only increases the recognition rate to some extent, but also has the advantages of building model quickly and high generalization ability of NFS.

Key words: AdaBoost.M2, Neural-Fuzzy System(NFS), plant recognition, Support Vector Machine(SVM), membership