3D Pose Estimation of Porous Workpiece Basedon Binocular Vision System

The workpiece with multi-hole is common in the field of machine production and manufacturing. Robot-oriented automated docking is an important means to improve product assembly efficiency. How to accurately estimate the position and the pose of porous workpiece is drawing more and more attention. In this paper, we propose a new binocular vision method for 3D pose estimation. First, an ellipse detection method is used to extract the hole feature. It uses three geometry constraints to ensure the accuracy of the ellipse detection. Second, the vanishing point and vanishing line of the workpiece plane are calculated using the ellipse information. Then the real projection point of the circle center can be accurately obtained through the vanishing line. The position of the circle center is reconstructed by the triangulation technology in our method. Last, the structural characteristics of the multi-hole workpiece are fully employed in the step of pose estimation. The posture is calculated by fitting the plane of those 3D points of circle center in order to reduce the error. Experiments are performed. The precision error of position estimation could be within 0.15mm and the precision error of pose estimation could be within 0.5°, meeting the operation requirements of machine production and manufacturing. These results verified the high accuracy and strong robustness of our proposed method.