移动机器人视觉传感器的现场标定技术

提出一种实用的移动机器人双目视觉传感器的标定技术。为解决非线性标定过程中解的不稳定性,根据透视投影的交比不变性和透镜成像的径向排列约束(radial parallelism constraint,RAC),线性标定摄像机的内部参数。然后,假定内部参数不变,从双目成像机理出发推导出本质矩阵的分解形式,线性标定出旋转矩阵和与实际只相差1个比例因子的平移向量。最后,由已知两个控制点之间的距离,求出实际的平移向量。实验表明,该方法具有与传统方法相同的精度,并且可实现外参数的现场标定。

On-line Calibration Technique of Vision Sensor in Moving Vehicles

A practical approach for stereo calibration in moving vehicles is proposed.To solve the instability of solution in the nonlinear calibration,intrinsic parameters are computed using cross-ratio invariability and Radial Parallelism Constraint(RAC) linearly.Then,assuming that intrinsic parameters are invariable,the deduction on decomposition form of essential matrix based on stereo imaging principle is given,and it can calibrate rotation matrix R and translation matrix T up to a scale factor linearly.Finally,the real extrinsic parameters are estimated with knowledge of distance between two scene points.Experimental results show that this method has same accuracy as that of traditional methods and can calibrate extrinsic parameters in the field.