提出了一种在复杂背景下,无需人工干预、自动提取二维正弦标靶特征点的方法。以液晶显示器作为平面标靶,标定图像为正弦强度条纹。摄像机拍摄标定条纹图后,算法根据条纹调制度自动分离背景和有效条纹区域;先将水平和垂直相移条纹的截断相位进行相位展开,再将两个相位分布构造成新的二维正交条纹图,通过多项式拟合建立各相位分布与其像素坐标的关系,以正交条纹亮点(对应相位为2π)作为标定特征点,并对这些亮点进行定位操作,获取其对应的亚像素坐标;最后利用展开相位的级次与条纹周期信息自动计算出它们的世界坐标,用于后续的摄像机标定。与常用的二维棋盘格对比实验结果表明,该方法标定结果可靠,而且精度更高,最为关键的是,该方法无需人机交互的手动操作,自动易行,可在以二维正弦条纹为标靶的摄像机标定中推广应用。

A feature-point extraction method in camera calibration without human intervention even in complex situation is proposed in this paper. The calibration target is an LCD screen by displaying the sinusoidal fringe patterns. The range of interest can be automatically segmented on the base of the fringe modulation. The two orthogonal phase distributions got by the phase-shift method can further create a cross-fringe pattern, in which the local highlight points with 2π phase value are chosen as the feature points for camera calibration. The sub-pixel image coordinates of those feature points can be got by polynomial fitting between phases and image coordinates. The world coordinates can be calculated by fringe orders and fringe periods directly. Compared with commonly used checkerboard calibration, the proposed method is reliable and has higher accuracy. Moreover the method is automatic, easy to use and without any human intervention. It can be helpful for the application of the cross-fringe patterns in camera calibration.