摘要
为提高基于工业机器人近距离下空间平面位姿的求解精度,分析了由透视畸变引起的圆投影误差的变化规律,以物距与特征圆半径的比值作为衡量圆透视畸变误差的参数依据,提供了不同物距下的精度参考。推导了基于空间圆投影图像的三点相对位姿求解算法。定量的误差分析和实验验证表明,物距与特征圆半径比值在18以内像素精度在0.1~1像素时,算法精度平均可以达到0.1°,优于文中其他两种算法。
Industrial robots are gradually applied in the field of precision measurement on account of its high degree of freedom and high precision of repetitive positioning.In order to achieve autonomous movement and complete specified actions,industrial robots often need to carry out precise motion control and precise measurement of the physical scale of the target.The former is accompanied by the development of the manipulator,and its control algorithm has become mature,while the latter has an increasingly high demand for relevant algorithms due to the changing actual working scenarios.As an important part of the autonomous motion of industrial robots,the solution method and efficiency of the spatial plane pose are directly related to the final motion accuracy of robots.In order to simplify the robot structure and improve the sensitivity of the whole system,the monocular vision system is adopted in this paper,which has a wider working range and a more compact structure.At present,the solution of the spatial plane pose can be divided into multi-point constraint problems and non-point constraint problem.The multi-point constraint problem usually involves the identification and pose settlement of multiple feature points,and its application often requires the use of a specific calibration board or the addition of multiple auxiliary feature objects,so the solution is complicated.The non-point constraint problem mainly solves the space pose of the plane by analyzing the characteristic circle.As a common basic feature in life,the circle appears widely on all kinds of workpiece.The research of the spatial plane pose algorithm based on the spatial circular feature is very important to improve the autonomous measurement and motion control of industrial robots.Because of camera perspective distortion,the image of the circle in the camera is not a standard ellipse,but a quasi-ellipse.There is no accurate basis for ignoring this perspective distortion to treat the projection of a circle as an ellipse.In this paper,by analyzing the variation of elliptic-like short diameter length in circle imaging,the ratio of object distance d to circle radius r is taken as the parameter to measure the distortion error of circle perspective,and its maximum value is taken as the precision value of each ratio of d to r,and the accuracy variation curve of perspective distortion is given.The traditional algorithms for solving spatial plane pose based on the characteristic circular target mainly involve the application at a long distance,and the solution shows duality,which often requires complex calculation.In this paper,considering perspective distortion,we mainly study the problem of using the feature circle to solve the position and pose of the space plane in a short distance.By using three points in the projection image of a feature circle to solve the relative position and pose of the target plane and the camera,we derive the algorithm of solving the relative position and pose.In order to verify the accuracy and actual effect of the algorithm,two comparison algorithms are added.By establishing an error analysis model,error simulation and experimental verification of the three algorithms are carried out respectively.The results show that the average accuracy of the algorithm can reach 0.1°when the ratio of object distance to feature circle radius is within 18 and the prime precision is 0.1~1 pixel,which is better than the other two algorithms in this paper.
作者
丁江
宋朝成
马翠
何凯
左启阳
DING Jiang;SONG Chaocheng;MA Cui;HE Kai;ZUO Qiyang(College of Mechanical Engineering,Guangxi University,Nanning 530004,China;Shenzhen Institute of Advanced Technology,Chinese Academy of Sciences,Shenzhen 518055,China)
出处
《光子学报》
EI
CAS
CSCD
北大核心
2023年第2期237-249,共13页
Acta Photonica Sinica
基金
NSFC‒深圳机器人基础研究中心项目(No.U2013204)
国家自然科学基金(No.51905113)
广西自然科学基金(No.2021GXNSFAA220095)。
关键词
测量
机器视觉
空间圆特征
位姿测量
透视投影
Measurement
Machine vision
Spatial circular feature
Pose measurement
Perspective projection
