三爪卡盘的亚像素边缘检测与视觉定位

Subpixel edge detection and vision location of three-jaw chuck

针对面向数控机床三爪卡盘的智能视觉上、下料机器人,对识别和定位精度要求较高,提出基于三爪卡盘形心特征的高斯二次曲线拟合亚像素的边缘定位方法,完成机器人对于三爪卡盘的识别及定位操作。其步骤为:首先,应用Canny边缘检测算法对目标边缘进行粗定位,即像素级定位;其次,建立三爪卡盘内切圆和外接圆组成的形心特征约束,根据边缘所在位置邻域像素的几何特征和灰度分布特征,对亚像素级别进行定位,使边缘定位在更加准确的位置。研究结果表明:机器人对于三爪卡盘的视觉测量精度达到0.01 mm,视觉定位精度达到0.05mm,达到数控加工行业精度要求;该技术在保证精度的同时,大幅度地提高了运算速度。

Based on the fact that in the clamping robot for three jaw chuck in numerical control（NC） machine tools,the clamping robot with vision system has high demand for recognition and position,a novel subpixel location method in which the Gaussian quadratic curve fitting is under the centroid guidance was put forward.The procedures were as follows.Firstly,the rough edge,as the pixel level edge,was achieved by the Canny algorithm.Secondly,the centroid determined by the centre of incircle and circumcircle in three-jaw chuck was established.Thirdly,referring to the geometrical characteristics and the gray distribution characteristics of neighborhood pixels,the subpixel level location was completed under the centroid guidance.At last,the experiment was made.The results show that vision measuring accuracy reaches 0.01 mm,and positioning accuracy reaches 0.05 mm,which satisfies the NC machining requirements. This technology guarantees machining accuracy,and greatly improves the computational efficiency.