Augmented Reality Geographic Information System (ARGIS) applications can only provide users accurate content services with a highly precise geo-registration. However, the absolute 6DOF (Degree of Freedom) pose provided by the portable sensors is usually inaccurate in urban outdoors, resulting in poorly geo-registration accuracy for ARGIS applications. Aiming at this issue, an automatic vision-aided localization method based on the 2D map is proposed to improve the initial localization accuracy of the portable sensors, and an overall geo-registration optimization framework for outdoor ARGIS is proposed. Based on the initial pose provided by the sensors, the basic principles of the vision-aided localization method are expounded in detail. The experimental results show that the proposed method can effectively correct the initial pose obtained by the pose sensors, and improve the geo-registration accuracy of outdoor ARGIS applications ultimately.