Abstract: SVM has been successfully employed to solve classification of remote sensing images.Total margin-based Adaptive Fuzzy Support Vector Machine（TAFSVM） which has good quality is proposed.TAFSVM not only solves the overfitting problem resulted from the outliers with the approach of fuzzication of the penalty，but also corrects the skew of the optimal separating hyperplane dut to the very imbalanced data sets by using different cost algorithms.In addition，by introducting the total margin algorithm to replace the conventional soft margin algorithms，a lower generalization error bound can be obtained.Besides，real-valued genetic algorithms optimize its parameters.Subsequently，AGATAFSVM is used to classify the data of remote sensing images.The experimental results indicate that the proposed AGATAFSVM can achieve higher classification accuracy and is stabler than radial basis functions neural network，K-nearest neighbors classifier and standard SVM.