Abstract: A pLSA based Topo-Markov random field(MRF) model method for Synthetic Aperture Radar(SAR) image classification is proposed in this paper.A Topo-category learning method is proposed here to represent the relationships by calculating the proportions of points on boundaries to points belong to each class.It has superiorities over consistent quadratic terms as Potts models and complicated quadratic terms.Meanwhile,Local Binary Pattern as well as other typical features is used as the candidates of the input with feature selection strategy.The experiments are carried on the MSRC and SAR image datasets and the results reveal the proposed algorithm's efficient performances and superiorities.