为了在光切法三维形貌测量中快速准确提取光条的中心位置，保证测量精度，研究了一种光条中心的亚像素提取方法。将图像的灰度视为关于坐标的二维函数，在CCD拍摄的激光光条图像中求取灰度的偏导数和梯度，图像中每一点的梯度代表此点的灰度变化方向，将此方向简化为光条上的点的法线方向。对梯度的大小设定阈值并结合该点的灰度信息，提取图像灰度变化最激烈的部分，即光条饱和带的像素位置。对光条饱和带的每个像素，按照不重复计算的原则在梯度方向上和规定邻域内求灰度中心，得到精确到亚像素级的光条中心位置。此方法将光条曲线的法线方向简化为灰度的梯度方向，省去了求海赛矩阵以确定法向的步骤。实验证明，同等条件下，该方法的计算耗时为常规亚像素提取方法的10%。利用该算法进行实际测量，在200 mm×200 mm测量范围下，误差(3σ)为0.057 mm，满足了测量对精确性和实时性的要求。

In order to extract the central positions of a light stripe and to ensure the accuracy of surface measurement based on light-sectioning method, an approach of the light stripe central position extraction in sub-pixel is investigated. By taking the gray level of an image as a 2D function of coordinates, the directional derivative and the gradient of the image of light stripe are calculated.In the image, the pixel gradient represents the direction of the change in gray level, and it is assumed as the normal direction of the light stripe curvilinear. Pixels at the positions of light stripe can be extracted by setting a threshold for both absolute values of the gradient and the gray level. The sub-pixel central positions of the light stripe are equal to the mean value of the gray-level-weighed positions of the pixels in the stripe’s adjacent region along the gradient direction. With this method, the normal direction of the stripe can be simplified to be the gray level gradient, which avoids complicated steps to calculate a Hessian matrix to get normal directions. Experimental results indicate that the elapsed time of the method is only 10% that of the traditional way, and the error (3σ) of real measurement is 0.057 m in the range of 200 mm×200 mm. It can satisfy the requirements of accuracy and real-time measurement.