Aiming at the problem of autonomous navigation and positioning under the circumstance of GPS signal rejection in wartime, a visual navigation algorithm based on global positioning reference map is proposed. Firstly, we map the scene by SLAM algorithm in non-wartime. To obtain the global positioning reference map, we align the global positioning data of RTK (real-time kinematic) by using the time stamp secondly. At last, we load the global positioning reference map by online SLAM algorithm, and the global navigation and positioning is running at the same time. The algorithm, which is able to complete the global positioning only according to the visual image data, is suitable for the applications of vehicle autonomous navigation. The result of the experiment shows that the accuracy of positioning and navigation reaches the level of decimeter. That is, the average error of navigation and positioning and the mean square deviation are 0.36 m and standard deviation is 0.31 m, which meets the requirements of practical applications.