面向视觉伺服的工业机器人轮廓曲线角点识别

Corner recognition of industrial robot contour curve for visual servoing

为了解决工业机器人视觉伺服中轮廓曲线角点识别效率低进而影响实时定位准确性的问题,提出面向视觉伺服的工业机器人轮廓曲线角点识别算法.建立基于Freeman链码的机械爪轮廓曲线编码模型,利用差别码对机械爪轮廓曲线的角度变化特征进行统一编码建模.通过基于差别码和卷积系数的卷积运算,对差别码进行局部非线性加权,计算机械爪轮廓曲线上各点的曲度精确量化辨识机械爪角点.基于候选角点阈值、点的最远领域间隔和局部范围最多点数进行获选角点的初步选取.采用平面轮廓曲线毛刺过滤筛选误识角点,实现机械爪角点的精确识别,为视觉伺服中实时定位提供可靠的位置信息.通过与已有平面轮廓曲线角点识别算法对比,验证本研究所提方法的准确性和高效性,提取角点过程具有较强的鲁棒性.

A contour curve corner recognition algorithm was presented to solve the problem of low efficiency of corner recognition of contour curve in visual servoing of industrial robots,which affects the accuracy of real-time positioning.A coding model of gripper contour curve based on Freeman chain code is established,and the differential code is used to unify the coding model of the characteristics of contour curve angle change of gripper.The differential codes are weighted locally and the curvature values of points on the gripper contour curve are calculated to accurately quantify the identification of the corners through convolution operation based on differential codes and convolution coefficients.The preliminary selection of the selected corner points is based on the candidate corner point threshold,the distance between the farthest point of the point and the maximum point value in the local range.A burr filtering for planar contour curve is used to screen mistaken points,moreover an accurate recognition of gripper corner point is realized,which provides reliable location information for real-time positioning in visual servo system.The accuracy and efficiency of the proposed method are verified by comparing with the existing corner recognition algorithms of planar contour curves,and the corner extraction process of the proposed method has strong robustness.