Ceiling vision localization with feature pairs for home service robots

Abstract

Automatically self-localization of home service robot in indoor environments is a key issue with high efficiency and robustness. State-of-the-art frameworks in computer vision typically are based on Simultaneous Localization And Mapping (SLAM) and extended version such as Ceiling Vision SLAM (CV-SLAM). However, a large amount of map information about landmarks in the ceiling are redundant for home service robots only to accomplish the localization task such as indoor cleaning robot. In this paper, a fast and robust ceiling vision localization framework based on CV-SLAM is proposed, which consists of three parts: ceiling features operation, localization with feature pairs and visual odometry. In addition, we propose a novel localization method with feature pairs based on ceiling vision, which can enhance the real-time capability effectively. Simulation and real experiments validate the efficiency of our approach for the localization tasks of home service robots in indoor environments.