3D Map Registration using Vision/Laser and Inertial Sensing.

摘要

Abstract — A point cloud registration method is proposed in this article, and experimental results are presented for long threedimensional map sequences obtained from a moving observer. In vision based systems used in mobile robotics the perception of self-motion and the structure of the environment is essential. Inertial and earth field magnetic pose sensors can provide valuable data about camera ego-motion, as well as absolute references for the orientation of scene structure and features. In this work we explore the fusion of inertial and magnetic sensor data with range sensing devices. Orientation measurements from the inertial system are used to rotate the obtained 3D maps into a common orientation, compensating the rotational movement. Then, image correspondences are used to find the remaining translation. Results are presented using both a stereo camera and a laser range finder as the ranging device. The laser range finder also needs a single camera to stabilish pixel correspondence. The article overviews the camera-inertial and camera-laser calibration processes used. The map registration approach is presented and validated with experimental results on indoor and outdoor environments. Index Terms — Inertial Sensors, 3D Mapping, Laser Sensors I.