Abstract: Based on a linear pinhole camera imaging model, the crass-ratio invariant has been applied to pattern recognition on computer vision for a long time as the most fundamental projective invariant. Then, the error of cross-ratio invariant is proposed and studied. Analyzing the influence of image noise on cross-ratio accuracy, the upper limit and lower limit of error are both obtained. According to the cross-ratio invariance, a new well-simplified artificial landmark pattern is designed to provide very high robustness for detecting under all kinds of viewing angles and illuminant conditions. The detection results of real images showed that the new artificial landmark pattern based on cross-ratio invariant is easy to detect fast and accurately. Landmark pattern is of practical importance in application to the self-localization and the navigation of a single-view autonomous mobile robot based on artificial landmark.