基于NURBS曲线拟合的图像亚像素边缘提取方法

通过介绍激光扫描测量原理 ,说明准确地从光刀图像中提取光刀边缘位置是影响测量精度的关键因素 .由于噪声的存在 ,使用常见的边缘检测方法 ,在光刀图像中提取同一位置的光刀边缘 ,以及计算光刀中心位置 ,会带来很大的误差 .为了降低误差 ,本论文提出了基于 NURBS曲线拟合的亚像素光刀边缘提取方法 .通过实验 ,说明该方法重复提取的结果比较稳定 ,而且将误差控制在 0 .4个像素以下 ,从而使重复测量的高度误差在 0 .0 8mm以下

Method to Image Edge Extracting Based on NURBS Fitting

This paper argues that how to compute the center position of the knife in the image obtained by CCD accurately is a key issue by introducing the principle of laser scanning system. Because of the influence of noises, great errors can be brought into the system when the common methods are used in computing the same position in many times. In order to reduce errors, this paper designs a new method based on NURBS curve fitting. By experiments, the results computed by this method are stable and the errors is controlled below 0.4 pixel. Thus, the errors in the height direction are less than 0.08mm.