CAD-Based Pose Estimation for Random Bin-Picking of Multiple Objects Using a RGB-D Camera

Abstract

In this paper, we propose a CAD-based 6-DOF pose estimation design for random bin-picking of multiple different objects using a Kinect RGB-D sensor. 3D CAD models of objects are constructed via a virtual camera, which generates a point cloud database for object recognition and pose estimation. A voxel grid filter is suggested to reduce the number of 3D point cloud of objects for reducing computing time of pose estimation. A voting-scheme method was adopted for the 6-DOF pose estimation a swell as object recognition of different type objects in the bin. Furthermore, an outlier filter is designed to filter out bad matching poses and occluded ones, so that the robot arm always picks up the upper object in the bin to increase pick up success rate. A series of experiments on a Kuka 6-axis robot revels that the proposed system works satisfactorily to pick up all random objects in the bin. The average recognition rate of three different type objects is 93.9% and the pickup success rate is 89.7%.