• 论文 •    

无标定摄像机手眼系统平移下的目标深度估计

孔令富,吴培良,李贤善   

  1. 燕山大学 信息科学与工程学院,河北秦皇岛066004
  • 出版日期:2009-08-15 发布日期:2009-08-25

Object depth estimation using translations of hand-eye system with uncalibrated camera

KONG Ling-fu,WU Pei-liang,LI Xian-shan   

  1. College of Information Science&Engineering, Yanshan University, Qinhuangdao 066004, China
  • Online:2009-08-15 Published:2009-08-25

摘要: 在基于图像的视觉伺服手眼系统中,往往需要通过估计获得目标的深度信息。为此,提出了通过控制装有摄像机的机械手进行纯平移运动,从获取图像序列特征点的变化来估计目标深度的新方法。首先利用四次纯平移运动,离线计算手眼关系的旋转矩阵;在线阶段,利用该旋转矩阵将机械手平移转换为摄像机平移,再结合平移前后图像极点不变的特性,实现目标深度实时估计。与以往方法不同,新方法无需进行摄像机参数标定,并对图像匹配噪声和手眼系统运动误差具有一定的鲁棒性。

关键词: 深度估计, 手眼系统, 平行移动, 无标定摄像机

Abstract: In the hand-eye system by image-based visual servo, depth information of the object was usually obtained by estimation. To deal with this problem, a new depth estimation method with hand-eye system was proposed, which was based on pure translations of manipulator fixed with a camera and shifts of keypoints in the images captured before and after camera translation. Firstly four pure translations were executed offline to obtain the rotation matrix of hand-eye relationship, At the on-line stage the translations of camera were calculated according to the translations of manipulator and the rotation matrix of hand-eye relationship, and the depth of the object was estimated in real time utilizing particularity which the epipolar was steady under pure translations. Different from other methods, this method needed not to calibrate the parameters of camera, and was robust enough to resist image noise and mechanical movement noise.

Key words: depth estimation, hand-eye system, translation motion, uncalibrated camera

中图分类号: