Abstract: This paper presented a novel affine invariant image matching algorithm based on subspace theory. Firstly, the nonlinear geometry structure of the affine invariant feature space was explored and the affine invariant subspace image features were extracted; then the global consistent matching of shape feature was realized based on the theory that the feature matrix of target shape can be determined only by its orthogonal projection matrix; finally, the same target images were matched with affine transform through computing the subspace distance. Matching experiment results on simulated transformed real images and real images shows that the proposed algorithm exhibits higher capacity to affine transform and robust to perturbations.