Abstract: This paper presents a new geometrical method which use 4 coplanar points and their correspondences on image plane to compute the camera pose considering the lens distortion of radial alignment constraints. It also use random sample consensus technique and intrinsic parameters consistency constraints to improve the stability and accuracy of the results. The imperfectness of camera pose assessment by projection test and back-projection test is exposed. We suggest that the motion transformation between different view points should be used as important clue to verify the corresponding camera pose data.