基于共面圆的距离传感器与相机的相对位姿标定

近年来, 距离传感器与摄像机的组合系统标定在无人车环境感知中得到了广泛的研究与应用, 其中基于平面特征的方法简单易行而被广泛采用. 然而, 目前多数方法基于点匹配进行, 易错且鲁棒性较低. 本文提出了一种基于共面圆的距离传感器与相机的组合系统相对位姿估计方法. 该方法使用含有两个共面圆的标定板, 可以获取相机与标定板间的位姿, 以及距离传感器与标定板间的位姿. 此外, 移动标定板获取多组数据, 根据计算得到两个共面圆的圆心在距离传感器和相机下的坐标, 优化重投影误差与3D对应点之间的误差, 得到距离传感器与相机之间的位姿关系. 该方法不需要进行特征点的匹配, 利用射影不变性来获取相机与三维距离传感器的位姿. 仿真实验与真实数据实验结果表明, 本方法对噪声有较强的鲁棒性, 得到了精确的结果.

Relative Pose Calibration Between a Range Sensor and a Camera Using Two Coplanar Circles

Relative pose calibration between a range sensor and a camera has been widely studied and applied in the environment perception of unmanned vehicles, of which the methods based on the planar features are greatly easy to be implemented. However, most of the current methods are based on point matching, which is easy to have errors and low robustness. In this paper, a relative pose calibration method between a range sensor and a camera from two coplanar circles is proposed. Using such a calibration object including two coplanar circles, the pose between the camera and the calibration object can be determined as well as the pose between the range sensor and the calibration object. Moreover, moving the calibration object to obtain more data, the center coordinates of two circles in the range sensor and camera coordinate system can be computed to refine the reprojection errors and 3D-3D correspondence point errors. Then, the pose between a range sensor and a camera can be estimated. The advantages of this method are as follows: matching among multiple points are not needed by using projective invariance. The simulation and real data experiments proved that this method has high accuracy and robustness.