An investigation on the accuracy and the repeatability of a designed pose measurement system based on the stereo vision

Recently machine vision technology is employed in many industrial measurement systems such as estimating end-effector pose of mechanical manipulators, which is a critical process in the kinematic calibration or feedback control of manipulators. Due to this fact that the accuracy and repeatability are two essential properties of any desired measurement system, in this article we examined and reported the accuracy and repeatability of a stereo vision 3D pose measurement system, which has been designed and implemented in Matlab software in order to estimate the pose of a parallel robot end-effector. Experiments were conducted in the defined depths from a 640*480 resolution stereo rig and the results were compared with another measurement system including 0.1µm precise gauge blocks. The results show the acceptable accuracy of the system; 1.2775mm for Z=300mm and 1.9297mm for Z=500mm. the repeatability of the system is about 1.5mm for both distance. However more accurate results can be achieved by improving the stereo rig. Moreover, criticality of camera calibration procedure in vision-based measurement systems is proved.