摘要
结合视觉的工业机器人轮毂加工过程中普遍存在轮毂工件识别效率低、定位不准确的问题,针对此问题提出一种基于机器视觉的轮毂识别定位方法。该方法利用工业机器人搭载工业相机采集工件台上轮毂图像,对其经过背景去噪和局部线性增强预处理后,采用限制位姿区域的虚拟球面视图法获取轮毂模板视图,再结合图像金字塔分层模板与归一化余弦相似性原理进行搜索匹配,从而初步解算出轮毂在相机坐标系下的六自由度位姿,并使用Levenberg-Marquardt法对位姿结果精度优化。通过开展大角度倾斜视角、部分遮挡、正常光照、局部反光、阴影多个场景下的轮毂识别与定位实验,结果表明该方法识别准确、定位精度高,轮毂定位测量平均误差为1.0551 mm,标准差为0.3467 mm,为工业机器人在实际加工中实现轮毂精确识别与定位提供了有效且鲁棒的解决方案。
Addressing the common issues of low recognition efficiency and inaccurate positioning of wheel hub workpieces in the machining process of vision-based industrial robots,a machine vision-based method for wheel hub identification and positioning is proposed;the method utilizes an industrial robot equipped with an industrial camera to capture wheel hub images on the workpiece table,and after background noise removal and local linear enhancement preprocessing,employs a restricted-pose virtual spherical view method to obtain the wheel hub template view,then uses layered templates from an image pyramid combined with the normalized cosine similarity principle for search matching to initially compute the wheel hub′s six-degrees-of-freedom pose in the camera coordinate system,with the pose further refined using the Levenberg-Marquardt algorithm;experiments under various scenarios-including large-angle oblique views,partial occlusions,normal illumination,local reflections,and shadows-demonstrate that the proposed method achieves accurate recognition and high positioning precision,with an average positioning error of 1.0551 mm[JP]and a standard deviation of 0.3467 mm,thereby providing an effective and robust solution for precise wheel hub identification and positioning in industrial robot machining.
作者
赵怀宇
刁燕
段必成
罗华
ZHAO Huaiyu;DIAO Yan;DUAN Bicheng;LUO Hua(School of Mechanical Engineering,Sichuan University,Chengdu 610065,China)
出处
《组合机床与自动化加工技术》
北大核心
2026年第2期17-20,26,共5页
Modular Machine Tool & Automatic Manufacturing Technique
基金
四川大学-宜宾市校市战略合作专项资金项目(2020CDYB-17)
四川省科技计划项目(2022YFG0220,2024YFFK0133)
四川省科技厅项目(2020KJT0117-2020YFQ0039)。
关键词
机器视觉
轮毂定位
模板匹配
位姿精度优化
machine vision
wheel hub positioning
template matching
pose accuracy optimization
