基于半镜面靶标的无重叠视域相机全局标定

A flexible external parameter calibration for non-overlapping cameras based on mirror reflect collinearity

摘要:

针对智能无人车、监控等多相机全局标定问题，往往存在无重叠视域或无法单向观测的情况，因此本文提出了一种基于半镜面平面靶标的多相机全局标定方法。该方法首先完成了单相机的内参数标定，然后固定点激光器并投射光线到第一个相机c1视场内并进行散斑点的观测，接着利用半镜面平面靶将光路反射到第二个相机c2视场内并观测散斑点像，由于单相机自身的内外参数已经标定，因此可计算得到反射光路的直线方程，最后根据直线的方向向量可计算全局坐标系的转换矩阵 和平移向量 。仿真和实验的结果表明，所提方法在500mm-1200mm距离，相机视域 范围内，其标定误差为0.36毫米。相对于现有方法，本文所提方法具有更广泛的适应性，可针对多相机复杂视场结构完成标定，同时具有设备简单、操作方便、计算简便等优点。

Abstract:

This paper presented an external parameter calibration method for multiple cameras with non-overlapping fields of view (FOVs) based on a laser ray reflected from a half of the mirror. This method is especially suitable for the condition that the laser beam cannot go through the FOV of both cameras. Firstly, the internal parameter calibration of single camera is completed. Spot laser projects laser beam into the FOVs of the first camera, and then across the FOV of the second camera by part mirror calibration panel reflection. The rotation matrix and the translation vector between the two cameras are computed by the reflecting co-linearity property of the laser beam. The proposed method does not require any complex equipment, but one spot laser and one planar target with part mirror area. Experiment results show that when the working distance of the cameras is about 500mm-1200mm and field-of-view is around 500mm*400mm, the camera calibration max error is about 0.36mm. Compared with the present approaches, the proposed method has the advantages of simple equipment, easy to operation, easy to calculation.