A structure-aware visual SLAM system

In this paper, a new visual SLAM system which incorporates the structure information of the environment is proposed. The proposed system incrementally builds the map data and extracts structure information from the map points using a clustering method. The extracted prior structure is then embedded in the system by means of a new representation of the structure together with the corresponding map optimization method which can be solved in a sparse manner. The structure-aware system is evaluated in both indoor and urban environments where planar structures exist and the experimental results show that our system achieves higher accuracy and robustness in the presence of prior knowledge.