Real-Time Indoor Mobile Robot Localization Based on the Recognition of Omni-Directional Artificial Landmarks

This paper proposes a real-time robust localization scheme for mobile robots in indoor environments, based on the recognition of omnidirectional artificial landmarks captured by a single onboard camera. Considering the need of omnidirectional recognition and the disturbance of lighting condition, we encode the landmark identity with nested circles in black and white. The recognition algorithm consists of a global and a local recognition layers. The global recognition is a fast overall recognition process, including light detection, image clustering, region of interest (ROI) extraction, and ROI identification. If the number of identified ROIs does not meet the requirement of the localization algorithm, the local recognition will process those unidentified ROIs through adaptive ROI expansion and template cutting. Based on the landmark recognition, the absolute position and orientation of the camera in the environment are estimated using the geometric mapping between the image and global frames. The proposed approach is tested via experiments in a real indoor environment, and the result reveals high localization robustness and consistency to the lighting condition.Copyright © 2014 by ASME