UV-SLAM: Unconstrained Line-Based SLAM Using Vanishing Points for Structural Mapping

In feature-based simultaneous localization and mapping (SLAM), line features complement the sparsity of point features, making it possible to map surrounding environment structure. Existing approaches utilizing have primarily employed a measurement model that uses re-projection. However, direction vectors used in 3D process cannot be corrected because employs only lines’ normal Plücker coordinate. As result, problems like degeneracy occur during solved. To tackle problem, this letter presents UV-SLAM, which is an unconstrained line-based SLAM using vanishing points for structural mapping. This focuses on regularities without any constraints, such as Manhattan world assumption. For this, we use can obtained from features. The difference between observation calculated through image estimation vector defined residual added cost function optimization-based SLAM. Furthermore, Fisher information matrix rank analysis, prove measurements guarantee unique solution. Finally, demonstrate accuracy quality are improved compared state-of-the-art algorithms public datasets.