A common method used to obtain 3D range data with a 2D laser range finder is to rotate the sensor. To combine the 2D range data obtained at different rotation angles into a common 3D coordinate frame, the axis of rotation rela- tive to the mirror center of the laser range finder should be known. This axis of rotation is a line in 3D space with four degrees of freedom. This paper describes a method for re- covering the parameters of this rotational axis, as well as the extrinsic calibration between the rotational axis and a camera. It simply requires scanning several planar checker- board patterns that are also imaged by a static camera. In particular, we use only correspondences between lines in the laser scans and planes in the camera images, which can be established easily even for non-visible lasers. Further- more, we show that such line-on-plane correspondences can be modelled as point-plane constraints, a problem studied in the field of robot kinematics. We use a numerical solution developed for such point-plane constraint prob- lems to obtain an initial estimate, which is then refined by a nonlinear minimization that minimizes the “line-of- sight” errors in the laser scans and the reprojection errors in the camera image. To validate our proposed method, we give experimental results using a LMS-100 mounted on a pan-tilt device in a nodding configuration.
File in questo prodotto:
Non ci sono file associati a questo prodotto.