基于变权重最小二乘法图像边缘特征提取算法的研究及应用

Research and Application of Edge Feature Extraction of Digital Image by Least Square Method with Weight Function

为了快速准确检测机器视觉中零件边缘,针对具有清晰边缘的中高精度复杂形状的薄板类机械零件背光图像,提出一种数字图像的亚像素边缘特征提取方法。该方法首先用Ramer算法对大量亚像素边缘坐标数据按照直线和圆弧等基本几何元素进行分段。然后改进最小二乘法,根据离散点到预拟合曲线的距离设置权重,消除离值点的影响,迭代拟合零件轮廓的圆弧和直线的算法。同时计算圆弧直径、圆度精度、中心距尺寸精度。齿轮泵中间体零件的检测结果表明,该方法能快速准确检测小尺寸平面孔系的孔距、孔径等几何量精度,该方法具有较高的可拓展性,为机器视觉精密测量进一步的研究和应用提供参考。

A rapid and accurate contours fitting method for backlight image of thin and complex mechanical parts w ith clear edge is presented. The large number of sub pixel edge coordinate data w as divided into the basic geometric elements such as line and arc segments using the Ramer algorithm. The least square method w as improved to eliminate the impact from the outliers based on distance betw een discrete points and the pre fitted curve. The arcs and straight lines is fitted robust,then circle diameter and center distance w as calculated. The measurement results of gear pump intermediate part show that,this method can used in geometry precision measurement of thin parts w ith many holes quickly and accurately. This method is instructive for the further research and application in vision measurement.