Monocular SLAM System Based On Panoramic Vision

Intelligent robots are able to work in complex environments with the capacities of self-organizing and self-planning. Navigation problem is a hot issue concerned in researching on such kind of robots. Its aim is to move purposely and do the job without aids. With supports of the project of National Natural Science Foundation of China, this paper aims to develop a new algorithm for navigating based on the Simultaneous Localization and Mapping (SLAM) algorithm to make the robots totally autonomous in the unknown but structured environments. The navigation techniques are firstly reviewed in this paper and then the SLAM problem is introduced, based on the analysis of the localization problem and the map building problem which are two key points in the navigation techniques, including its structure, characteristics, and categories and so on. As the bi-camera based SLAM has shortcomings as high calibration complexity and high cost, this article proposes Panoramic Vision Based Monocular SLAM System. This article introduces the arithmetic in detail including panoramic image unwrapping and correcting, feature extracting and matching, feature depth calculating, building of robot's motion model and observation model, state updating, map constructing and managing. At last in order to solve the difficulties of this arithmetic in practical applications, this paper proposes using navigation arithmetic based on image indexing to work with this arithmetic. The experiment results based on the "GRANDAR" robot verify the validity of the algorithm.