In order to enhance the fast and stable ability of target recognition of high resolution SAR images in a complex battlefield environment. Due to the changing characteristics of size, orientation, rotation, and the severe influence of strong clutter background, that cause serious influence for the ground target recognition. A method of affine 3D object model to two-dimensional plane is proposed, Cosine-Fourier moment and Rayleigh distribution CFAR detection are used to extract the moment and peak features, respectively. The model and analysis of target recognition are done by cascade combination classifier in high dimension and pose change without additional overhead of guidance and control systems. Finally, experimental results prove it accurate and believable.