With the continuous development of augmented reality technology research and the deepening of map and geographic information applications, an increasing number of scholars are trying to apply mobile augmented reality technology to map display enhancement and geographic information visualization. At present, research of augmented reality maps faces challenges. The first challenge is lack of a specific marker for map space elements; furthermore, a marker without a semantic description and location characteristics is only used for augmented reality registration. In addition, there is a lack of three-dimensional (3D) tracking registration and virtual reality fusion for planar maps, which is key to achieving spatial consistency and enhanced expression of augmented reality maps. To solve the above problems, a new user-defined hierarchical semantic marker in the map space that combines AR mark and QR code information is designed and proposed. The user-defined hierarchical semantic marker in the map space changes the fixed markup mode in the traditional augmented reality. It can represent the map elements’ spatial locations, semantic information, and multi-dimensional attribute information with flexible features and universality. At the same time, the marker can be used to assist the virtual reality fusion and model matching, and lay a foundation for the application of the augmented reality map. This paper introduces a city map user-defined hierarchical recognition method, explains its principles in detail, and expounds on the process of 3D registration of real scenes and virtual model in space. Applying the registration method based on computer vision, four types of coordinate systems, i.e., the real space coordinate system, the virtual space coordinate system, the viewpoint coordinate system, and the image plane coordinate system, are unified to realize seamless fusion of the real scene and virtual information; then, a virtual reality fusion algorithm for map models is proposed. The virtual models corresponding to the user-defined markers are seamlessly integrated with the map in reality. Finally, results of conducted experiments and analyses of map marker hierarchy recognition, map virtual-real information fusion, etc., reveal that the virtual-real fusion between the user-defined map marker and map algorithm in this study is effective. The marker contains a wealth of information, the virtual models are displayed accurately, and the effect of enhanced expression is achieved. An analysis of the efficiency and recognition rate of the user-defined map marker hierarchy recognition reveals that the map marker hierarchy recognition method proposed in this paper, with a high recognition accuracy and a degree of universality, is fast, and its recognition efficiency meets the real-time requirements of augmented reality.