Kinect V2 Camera Based Vision System for Robotic Grinding

Abstract

The vision system often plays a vital auxiliary role in the high-precision grinding process of the robot. This paper presents an eye-in-hand vision system which incorporates a depth camera on the six-axis robotics and scans from multiple angles to extract the information of workpieces to be ground. Contour and centroid extraction were applied for locating workpiece and planning path of fine scanning. After obtaining the multi-angle point cloud images, we reconstructed the point cloud of the workpiece surface through point cloud registration and segmentation. Harris corner detection was used to recognize bumps on the surface. Then grinding information was uniformly converted to the robot coordinate system through coordinate transformation. The experimental results show that this system can obtain comprehensive and precise grinding information, ensuring the smooth execution of robotic grinding.