SO-SLAM: Semantic Object SLAM With Scale Proportional and Symmetrical Texture Constraints

Object SLAM introduces the concept of objects into Simultaneous Localization and Mapping (SLAM) helps understand indoor scenes for mobile robots object-level interactive applications. The state-of-art object systems face challenges such as partial observations, occlusions, unobservable problems, limiting mapping accuracy robustness. This letter proposes a novel monocular Semantic (SO-SLAM) system that addresses introduction spatial constraints. We explore three representative constraints, including scale proportional constraint, symmetrical texture constraint plane supporting constraint. Based on these semantic we propose two new methods - more robust initialization method an orientation fine optimization method. have verified performance algorithm public datasets author-recorded robot dataset achieved significant improvement effects. will release code here. 1