【作者】 张海明； 杨洗陈； 窦以红； 刘立峰； 高贵；

【Author】 ZHANG Haiming,YANG Xichen,GAO Gui,DOU Yihong (Laser processing center,Tianjin Polytechnic University,Tianjin,300160,China)

【机构】 天津工业大学激光技术研究所；

【摘要】 由双摄像机组成的机器人立体视觉标定包括两个摄像机单独标定,两个摄像机相对关系标定及摄像机与机器人之间的手眼关系标定。Matlab自带的标定工具箱,功能不完整,且标定过程需要手动指定边界点,耗时多,易造成误差。本文设计了一种新的成本低廉的平面靶标,对Matlab标定工具箱进行二次开发,针对新靶标提出了一种全自动标定方法。标定时,LabVIEW同时对两个摄像机进行图像采集,经过预处理,调用Matlab对靶标中的特征点进行高精度提取和匹配,并根据ZhangZY的方法进行标定计算,得到两个摄像机的内外参数和双目摄像机的相对关系。手眼标定时,摄像机安装在机器人末端,移动机器人末端改变3次位置,系统能自动计算出手眼关系矩阵,整个过程不需人工干预,精度高。采用的优化算法,使标定过程可以实时实现,实验中,对每幅图像的标定时间小于1秒。文章对自动提取算法,标定点配准算法做了详细描述,对实验结果进行了误差分析,并与传统棋盘格靶标结果进行比较。经比较,新方法实现了全自动标定,具有更高的精度和更小的标定误差,有很好的实用价值。更多还原

【Abstract】 The calibration of robot stereo vision with two cameras includes single-camera calibration,the relationship calibration between stereo cameras and hand-eye calibration.Camera Calibration Toolbox for Matlab is not well integrated.It needs to manually mark the boundary points,which is wasting time and can cause errors.A new low cost and simple planar target and a fully automatic method for the new target are presented in this paper.Images are acquired by LabVIEW from the two cameras.After preprocessing,the calibration point are extracted and matched by Matlab.According to Zhang Z Y’s calibration method,inside and outside parameters and the relationship between the stereo cameras are calculated.The camera is mounted on the robot arm.Move robot to 3 positions,the system will automatically calculate hand-eye matrix.The whole process is without human’s action.Optimization algorithm makes calibration done in real time.In experiments calibrating each image is less than 1 second.Automatic extraction and calibration point registration algorithm are described in detail in this article.Compared with the traditional method,the new methods can achieve a fully automatic calibration,with higher accuracy and smaller calibration error.It is very practical and useful.更多还原