Eye-in-hand vision-based robotic bin-picking with active laser projection

Abstract

In this paper, an eye-in-hand vision-based robotic bin-picking system is proposed. The system can identify the pose of a plumbing part from a pile and grip it correctly. A monocular eye-in-hand camera and a laser projector are employed to reconstruct the 3-D point cloud of plumbing parts stacked together. The projection direction of the laser line projector is controlled to change in order to scan the pile of objects while the camera is observing. 3-D points can then be determined by the a priori known geometry between the camera and the laser line projector. To estimate the pose of an object, the iterative closest point (ICP) is employed to match the point clouds of the object and the model. The transformation between the object and the model can thus be determined. A computed closer point (CCP) approach is proposed to estimate the pose of an object since the deviation from the object to the model is initially large in nature. The proposed CCP approach combining with the ICP algorithm can improve the success rate and accuracy of point cloud matching. The proposed system has been validated by experiments with potential applications in production lines.