In order to solve the problems of selection and recognition for relative positioning targets in complicated terrain,a new selection and recognition algorithm based on double matching correction strategies and maximally stable external regions (MSER) is proposed.The MSER features of reference image and real-time image are extracted respectively.The feature regions are performed through ellipse fitting and regularization,and then relative positioning targets are adaptively selected according to the regional selection weight index.Using scale and affine invariant of MSER features,the matching MSER features between the two images are obtained based on cross correlation criterion.Double matching correction is performed by applying Random Sample Consensus (RANSAC) method and the position weight index to delete wrong matching features and to realize the accurate recognition for relative positioning targets.In view of the complex ground scene,the experiments demonstrate that the maximum relative error recognition probability of the method is 0.125 and the absolutely error recognition probability is 0.028.It can satisfy the relative positioning requirements with advantages of higher precision and rapid speed,as well as strong anti-jamming and stabilization.