一种高效的视觉导航摄像机标定方法

Efficient Calibration Method for Vision Navigation Camera

针对移动机器人视觉导航中的摄像机标定问题,提出一种高效的摄像机标定方法。考虑透镜畸变的影响,利用Hough变换提取矩形棋盘格的角点,求出摄像机参数初始值,运用Levenberg-Marquardt算法对初始解进行非线性迭代优化。通过重投影方法将求出的角点重新投影到矩形棋盘格像平面上,并与原提取的角点相比较,分析产生位置误差的原因。实验结果表明,该方法抗噪声能力强,对实验条件要求相对较低,标定精度满足移动机器人实际导航的要求。

Aiming at the camera calibration in vision navigation system of mobile robot, this paper proposes an efficient method of camera calibration. The radial distortion and tangential distortion is taken into account. Hough transform is applied to extract the checkboard comers, and initial values of camera parameters are obtained. Levenberg Marquardt algorithm is used for nonlinear iteration optimization. The comers are reprojected to the rectangular checkboard, the comers extracted are compared with Hough transform, and the causes of error are analyzed. The results show that this method of anti-noise ability is strong, and the experimental conditions are relatively low. The precision completely satisfies requirement of actual navigation, which is suitable for camera calibration of mobile robot.