A Monocular Visual Localization Algorithm for Large-Scale Indoor Environments through Matching a Prior Semantic Map

It is challenging for a visual SLAM system to keep long-term precise and robust localization ability in large-scale indoor environment since there low probability of the occurrence loop closure. Aiming solve this problem, we propose monocular algorithm environments through matching prior semantic map. In approach, line features certain objects observed by camera are extracted real time. A cost function proposed represent difference between matched preexisting After that, bundle adjustment model integrating object given optimize pose real-time Finally, test cases designed evaluate performance our which with information advance build map The results show that positioning accuracy method improved navigation.