一种基于双目视觉和Halcon的高效机器人手眼标定方法  

An efficient robot hand⁃eye calibration methodbased on binocular vision and Halcon

作　　者：程强[1,2] 黄河[1,2] 许静静[1,2] 李江晗 李迎[1,2] 张涛[1,2] CHENG Qiang;HUANG He;XU Jingjing;LI Jianghan;LI Ying;ZHANG Tao(Institute of Advanced Manufacturing and Intelligent Technology,Beijing University of Technology,Beijing 100124,China;Beijing Key Laboratory of Advanced Manufacturing Technology,Beijing University of Technology,Beijing 100124,China;School of Electrical and Information,Northeast Agriculture University,Harbin 150006,China)

机构地区：[1]北京工业大学先进制造与智能技术研究所,北京100124 [2]北京工业大学先进制造技术北京市重点实验室,北京100124 [3]东北农业大学电气与信息学院,黑龙江哈尔滨150006

出　　处：《现代电子技术》2023年第13期35-42,共8页Modern Electronics Technique

基　　金：国家自然科学基金资助项目(5197050489)。

摘　　要：针对机器人在标定过程中追求快速、简捷、高效的特点,提出一种基于双目视觉和Halcon的高效机器人手眼标定方法。基于张正友棋盘格标定法和OpenCV中亚像素角点检测算法完成对双目相机内外参数的获取,对比运用不同的立体匹配算法和三维重建生成具有三维坐标的立体空间点云,从而得到相机坐标系下目标的位置坐标,之后取机器人坐标下的10组末端位置坐标以及相机坐标系下对应的10组位置坐标后,借助Halcon视觉软件的vector_to_hom_mat3d算子求解出相机和机器人坐标系的转换矩阵,完成机器人手眼标定。经过实验验证,该标定方法平均误差为3.58 mm,满足一般机器人工作要求,并且相比传统的手眼标定无需借助复杂的标定工具,且计算过程简捷高效。In view of the robot′s pursuit of fast,simple and efficient characteristics in the calibration process,an efficient robot hand⁃eye calibration method based on binocular vision and Halcon is proposed.The acquisition of internal and external parameters of the binocular camera is completed on the basis of checkerboard calibration method proposed by Zhang zhengyou and sub⁃pixel corner detection algorithm in OpenCV.The stereo space point clouds with 3D coordinates,which are generated by different stereo matching algorithms and 3D reconstruction,are contrasted to get the coordinate of the target location under camera coordinate system.With the help of the vector_to_hom_mat3d operator of the Halcon vision software,the transformation matrix of the camera and the robot coordinate system is solved after 10 groups of terminal location coordinates under camera coordinate system and corresponding 10 groups of location coordinates are obtained.And the robot hand⁃eye calibration is completed.The results of experimental verification indicate that the average error of this calibration method is 3.58 mm,which can meet the working requirements of general robots.In comparison with the traditional hand⁃eye calibration,this calibration method does not need complex calibration tools,and its calculation process is simple and efficient.

关 键 词：机器人标定 手眼标定 双目视觉 相机标定 立体匹配 三维重建 实验分析 

分 类 号：TN911.73-34[电子电信—通信与信息系统] TP242.3[电子电信—信息与通信工程]