面向机器人抓取的双目视觉系统标定与目标位姿估计算法  

Binocular Vision System Calibration and Object Pose Estimation Method for Robert Fetching

作　　者：江士雄[1] 曹丹华[1] 吴裕斌[1] 

机构地区：[1]华中科技大学光学与电子信息学院,湖北武汉430074

出　　处：《仪表技术与传感器》2016年第11期110-113,共4页Instrument Technique and Sensor

摘　　要：在双目视觉抓取系统中,首要任务是解决相机标定以及目标位姿估计问题。针对工业应用场景,文中提出了一种基于灭点一致性约束的相机标定算法,并利用目标的双目图像立体匹配结果,通过改进迭代最近点算法实现目标位姿估计。实验结果表明:该系统能够有效地估计目标位姿,目标位姿估计的平移与角度误差均能满足抓取需求。The camera calibration and object pose estimation are the fundamental problem in binocular vision system for object fetching. In the industrial application scenarios, a camera calibration method based on the vanishing point consistency constraint was proposed in this paper. Binocular image stereo matching algorithm was applied. And the object pose was estimated by the improved iterative closest points algorithm. The experimental results show that the system can effectively estimate target position , target pose estimation of translational and angular error can meet the demand of grab.

关 键 词：相机标定 双目视觉系统 位姿估计 嵌入式平台 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置]