Cylindrical Pellet Pose Estimation in Clutter using a Single Robot Mounted Camera

摘要

Pose estimation of cylindrical pellet using a single camera-in-hand configuration of a robot is discussed in this paper. Approaches to estimate pose in both isolated and an occluded environment is discussed. The pellet contour from the segmented image of the scene was compared with contours in the database to ascertain the matching orientation. For occluded pellets, a multiple-view based pose recognition system is proposed. Later, the estimated pose was communicated to the robot to enable it to pick-up the pellet. This has been experimentally implemented for cylindrical pellets and the performance is discussed. The algorithm enables online pellet pose determination and pick-up using KUKA KR5 robot.