基于光偏振信息的摄像机标定方法研究

Research on the Camera Calibration Method Based on Light Polarization Information

摄像机内参数标定直接影响位姿测量精度。在自然光条件下采集图像,由于光照不均匀,使靶标图像产生局部高亮现象,降低了标定点图像坐标提取精度。针对这一问题,提出一种基于光偏振信息的摄像机标定方法,分两步完成摄像机标定,第1步,使用自然光条件下采集的图像完成摄像机标定;第2步,在镜头前加装偏振片,根据第1步得到的标定结果,针对不同偏振角条件下采集得到的图像,计算靶标平面与摄像机像平面之间的2D单应矩阵进而根据相应的重投影误差确定当靶标摆放在不同位置时所对应的最佳偏振角,最终使用该偏振角条件下采集的图像,完成摄像机标定。实验结果表明,该方法可以有效提升摄像机标定精度,进而提升位姿测量精度:在-45°~+45°范围内,平均测量误差低于0.019°,与使用自然光条件下标定参数的位姿测量相比,测量精度提升了0.01°。在0~20 mm范围内,平均测量误差低于0.03 mm,与使用自然光条件下标定参数的位姿测量相比,测量精度提升了0.043 mm。

The calibration of camera parameters direct affects the accuracy of pose measurement.When the image is collected under natural light conditions,due to the uneven illumination,the target image has a local highlight phenomenon,which reduces the extraction accuracy of the image coordinates of calibration points.To solve this problem,a camera calibration method is proposed based on light polarization information.The camera calibration is completed in two steps.Step 1,the images collected under natural light conditions are used to complete the camera calibration;Step 2,a polarizer is added in front of the lens,according to the calibration results obtained in the first step.For the images collected under the condition of different polarization angles,the 2D homography matrix between the target plane and the camera image plane is calculated,and then the optimal polarization angle is determined when the target is placed in different positions according to the corresponding re-projection error,and finally the images collected under the condition of polarization angles are used to complete the camera calibration.The experimental results show that the proposed method can effectively improve the camera calibration accuracy,and then improve the pose measurement accuracy.In the range of-45°-+45°,the average measurement error is less than 0.019°,and the measurement accuracy is improved by 0.01°compared with pose measurement using the calibration parameters under natural light conditions.In the range of 0-20 mm,the average measurement error is less than 0.03 mm,and the measurement accuracy is improved by 0.043 mm compared with pose measurement using the calibration parameters under natural light conditions.