天津职业技术师范大学学报

工业机器人在自动分拣装配中，当法兰盘有较大旋转角度时，使用普通模板匹配法对法兰盘进行定位时有较大误差。针对此问题，本文提出一种基于单目相机的定位方法，以某种法兰盘为例进行定位。对图像进行预处理及特征提取，得到各连通域的细长度；采用最小二乘法对法兰盘亚像素边缘进行圆拟合，得到法兰盘中心图像坐标并通过坐标变换将其转换到世界坐标系；使用2个长圆孔质心坐标对法兰盘旋转角度粗定位，采用改进的NCC模板匹配法对法兰盘的旋转角度进行精定位，得到法兰盘绕其主轴的旋转角度。实验结果表明：提出的方法旋转角度误差≤0.1°，抓取位置误差<0.3 mm，该方法在光照稳定的条件下可以对目标快速定位，并能够满足精度要求。

In the automatic sorting and assembly of industrial robots,when the flange has a large rotation angle,there is a great error in positioning the flange using the ordinary template matching method. For this issue,this paper proposes a positioning method based on a monocular camera,taking a type of flange as an example. Image preprocessing and feature extraction were carried out to obtain the elongatedness of each connected domain. The least square method was used to fit the sub-pixel edges of the flange to obtain the image coordinates of the flange center and convert them to the world coordinates system through coordinate transformation. The rotation angle of the flange was roughly positioned using the centroid coordinates of two oblong holes,and the improved NCC template matching method was used to precisely locate the rotation angle of the flange,thereby obtaining the rotation angle of the flange around its spindle. The experiment results show that the rotation angle error of the proposed method is not greater than 0.1°,and the grasping position error is smaller than 0.3 mm. Under the condition of stable illumination,the proposed method can quickly locate the target and meet the accuracy requirements.