Abstract
A vision-assisted positioning system is developed in this study. This vision-assisted positioning system is integrated in a PC-Based CNC controller. With the help of this system, the image of a workpiece on the work table is captured by a CCD camera, and the location of the positioning mark in machine coordinates is calculated. With these data, the translational and rotational relationship between workpiece coordinates and machine coordinates are established and returned to the coordinate management unit in the CNC control system. The actual deviations of the workpiece mounting location are compensated so that the CNC machine tools can perform the motion control and cut the workpiece with NC codes defined in workpiece coordinates. First, the working principle of the machine vision hardware has been researched to find the way and rules for selecting optimal components including camera, lens and lighting system. Different image processing methods are developed to analysis the captured mark and the accuracy of these methods are compared. In the end, a novel high accuracy method, the full-form grey function integration method, for the cross mark is used and integrated in a PC-Based CNC controller. The accuracy of the developed vision-assisted positioning system is successfully tested on a machine tool. The pixel resolution of the selected CCD camera is 40μm. The achieved average accuracy is -0.1836μm, repeatability for three standard deviations of the mark positioning is ±0.2694μm, which is about 1/70 pixel。