Fast localization and 3D mapping using an RGB-D sensor

Abstract

Low-cost range sensors represent an interesting class of sensors which are increasingly used for localization and mapping purposes in robotics. The combination of depth data and visual information can be employed to develop reliable algorithms for localization and environment mapping. A real-time approach combining a monocular visual odometry algorithm and range depth data is proposed in this paper. The scale factor problem is solved combining the depth data flow and the monocular image data. Moreover, a multiple resolution approach led by the distance of the sensor from surrounding obstacles is proposed for the depth data acquisition process. The visual egomotion estimation algorithm and the 3D map generation work in parallel improving the system realtime reliability. Experimental results show how the proposed integrated framework is able to localize in real-time the device in an unknown environment and to simultaneously generate an environment dense and colored map.