Based on the invariance of cross ratio, an effective method that employs more than three groups with four collinear feature points to compute accurate coordinates of the feature points falling on the light stripe plane is discussed,and the factors that may effect on the coordinates estimation accuracy are analyzed in detail. A solution to evaluate the uncertainty of constructing the feature points with invariant cross ratio is proposed. The simulation experiments reveal that the proposed approach is valid and feasible for the structured light vision sensor calibration. Furthermore, compared different methods to construct feature points with coordinate estimation uncertainty,the effective method to generate high accurate feature points is proposed, it is shown that the invariance of cross ratio can be applied more broadly in the filed of computer vision.