Visual Landmark Map-Based Spatial Recognition Using a Monocular Camera

In this paper, we propose a computer vision-based indoor spatial recognition method. The proposed method recognizes objects in a 360-degree space and eliminates errors in the process in which the space’s visual landmark map (VLM) is constructed by considering the 360-degree spatial arrangement and relative coordinates between the recognized objects. The method divides the target space into subspaces, and individual VLMs are applied to each subspace. The VLMs for these subspaces are activated based on their arrangement. When matching images from a camera with VLMs, the system identifies the user’s location within the space. An experiment demonstrated the effectiveness of the proposed method in accurately recognizing the user’s indoor position, and the data size for maintaining the VLM is highly efficient.