Uncalibrated monocular based simultaneous localization and mapping for indoor autonomous mobile robot navigation

This paper describes a an SLAM algorithm for the navigation for an indoor autonomous mobile robot. The main emphasis of this paper is on the ability of line extraction. A recognition method based on straight line extraction is proposed for extracting the key features on the office ceiling, in an effort to estimate the pose of mobile robot. Random sample consensus (RANSAC) paradigm is used to group the line segments. During the navigation, onboard odometry is used at the beginning stage to estimate the information of environment for visual reckoning, while lamps on the ceiling act as beacons for positioning to eliminate accumulation of errors after a long-term run. The data captured from infrared sensors is used for constructing a map. The proposed method scales well with respect to the size of the input image and the number and size of the shapes within the data. Moreover the algorithm is conceptually simple and easy to implement. Simulation and experimental results show that good recognition and localization can be achieved using the proposed method, allowing for the interested region correspondence matching and mapping between images from different sensors or the same sensor indifferent time phrase.