On the study of workpiece localization with a cost-effective vision based method

In a factory automation application, position measurement of a workpiece is crucial. Computer vision is one of the most important methods for localizing the workpiece. Special equipment, such as a camera and a laser range finder are often used to measure and detect the position of the workpiece. However, the process of obtaining these measurements are usually highly specialized and expensive. Our method concentrates on improving the accuracy of the workpiece position measurement without expensive equipment. We use Kinect as the 3D sensor, which consists of a RGB camera and a depth camera. First mark points in both the Kinect coordinate and the manipulator coordinate are obtained. Then calibration between Kinect and Manipulator is obtained using Quasi-Newton method with the initial mark points. In order to improve the accuracy, we propose an improved registration based on Iterative Closest Point (ICP) to further derive the final transformation matrix. Experiment shows that this comprehensive method is practical.