CV-SLAM using line and point features

Abstract

Simultaneous Localization and Mapping (SLAM) is a fundamental problem in the mobile robotics. As SLAM is usually utilized in an indoor environment, we select the ceiling view (CV) as a stable source of features. In this research, three types of features are extracted from CV and constitute a single map. The landmarks detected from ceilings are ceiling boundaries, electric lamps, and circles. Each of them is very robustly detected from CV and the combination of them yields more stable and accurate localization performance. Multiple kinds of features are integrated into an EKF-SLAM framework. We demonstrated the SLAM system in an indoor environment and proved its high performance.