Work pieces usually have burrs after machining processes. The burrs could result in a potential dimension error and they may cause the impression of following assembly processes. Therefore, the deburring process is very important for casting parts to guarantee the dimension accuracy and surface quality. Many factories still rely on the manual deburring for achieving a suitable quality of product. The reason is because the deburring machine can handle a specified part and it is not feasible for various parts. The manual deburring is highly time-consuming with danger. In addition, the resulting dust and physical demands ultimately represent a health risk. Most typical deburring techniques are applied for the fixed part and it is not easy to be automated. They require many setup processes as different positions are requested deburring for different parts. To improve the surface treatment of deburring parts, how to automate the deburring operation is a important issue in the development of manufacturing. In this study, to detect the burring size and position of a casting, a Realsense F200 camera is used to capture the 3D geometric model to establish a point cloud model. Then, a coherent point drift algorithm based on a variable ROI is proposed to detect burring dimension and location. After that, an automated deburring system is developed by using an UR3 robot integrated with the self-developed automatic optical inspection system.
point cloud, 3D vision, Coherent point drift algorithm, iterative closest point