Simultaneous Localization and Mapping (SLAM) is a process of building a map of an unknown environment and at the same time calculating current robot position. The aim of our investigation was to find a robust method to generate 3D maps with the use of latest sensor technologies like a RGB-D camera. We introduced 3D Visual SLAM framework by suggesting a system configuration for localization method, integrated with keyframe-based approach to accommodate need for backend processing. Loop closing system and post graph optimization were defined. This framework produces impressive results compared with other state of the art algorithms. The applications for our 3D Visual SLAM include MAVs vision navigation, environment perception and vision-based control system.
↧