In order to solve the problems of differences in scale,rotation,grayscale and 3D viewpoint,and achieve robust scene matching during electro-optical imaging terminal guidance,a new feature descriptor based on uniform patterns is constructed and a new approach based on the descriptor and maximally stable extremal regions (MSER) is proposed. It is steps of realizing the method that,the MSER features of reference image and real-time image which are scale and affine invariant are extracted respectively;rotation and grayscale invariant description based on uniform patterns are performed. The matching MSER features between the two images are obtained based on Euclidean distance ratio criterion;epipolar geometry of the two images is estimated by applying random sample consensus (RANSAC) method to the centers of gravity of the matching MSER features. Simulation results show that the proposed method provides robust scene matching during electro-optical imaging terminal guidance and is more robust than traditional methods.