足球机器人全局视觉系统的快速标定方法

A Fast Calibration Methold Applied in Global Vision System of Soccer Robot

摘要:

针对足球机器人全局视觉系统的特殊应用场景,提出了一种基于两步法简化标定模型的快速有效标定算法。利用近似欧拉角旋转矩阵建立摄像机标定模型,确定需要求取的摄像机内外参数,借助场地上现有的一些标志点作为靶标,对模型的相关参数进行分步求取初始值,并利用L-M优化算法对摄像机内外参数进行分步优化求取精确值。通过实验分别从标定误差和机器人动作实现效率两个方面与其他标定算法进行对比,验证了该算法具有较高的标定精度。实验表明,本算法的优点在于不用借助专门的标定靶标,即可快速有效完成相机标定,标定精度可达1mm,且标定过程简单易实现。

Abstract:

For the special application in global vision system of soccer robot, a fast and effective calibration algorithm was proposed based on two-stage simplified calibration model. Approximate Euler rotating matrix was used to build camera calibration model with the intrinsic and extrinsic parameters to be achieved. Some mark points on the field were used as reference targets to obtain the initial value of parameters step by step, and L-M method was adopted to optimize all the parameters step by step. Higher calibration precision was proved comparing with other calibration methods from two aspects of calibration error and robot movement realization by experiments. It was shown that by this algorithm the calibration can be realized easily and effectively with 1mm calibration preicsion without special reference target.