Auto-estimation of Relative Orientation Between a Roof-Mounted Spherical Camera and a Vehicle Based on Spherical SLAM

摘要

While humans can drive a vehicle using visual information, we investigate the possibility of driving assistance by using a single spherical camera. As the first step, this paper proposes a method of auto-estimation of the relative orientation between a roof-mounted spherical camera and a vehicle. The location of the spherical camera is estimated by a full-view SLAM (Simultaneously Localization and mapping) method. If the trajectory of the camera is on a plane, the normal vector of the plane can be computed by fitting the plane. By combining the heading direction of the vehicle with the computed normal vector, the coordinate system of the vehicle can be determined. Experimental results are given to show the effectiveness of the proposed method.