SLP-SLAM: A Multifeature Fusion Stereo SLAM Based on Structural Lines and Structural Planes

Accurate positioning and pose estimation of mobile robots in structured environments have persistently presented considerable challenges. This paper proposes SLP-SLAM, a multifeature fusion stereo simultaneous localization and mapping (SLAM) system that combines both structural lines and structural planes. We have enhanced the effectiveness of the method for extracting structural lines using the Manhattan World Assumption (MWA) and have suggested utilizing these lines to construct structural planes, which is different from the existing plane extraction methods. Furthermore, we develop the state estimator by minimizing the visual residuals of points, structural lines, and structural planes. As far as we know, this represents the inaugural stereo SLAM that concurrently employs points, structural lines, and structural planes for integrated optimization. An appraisal of our system on difficult public datasets has been conducted. The trajectory error associated with SLP-SLAM has been diminished by 15% in comparison to contemporary state-of-the-art (SOTA) SLAM.