Implementation of computer vision guided peg-hole insertion task performed by robot through LabVIEW

Abstract

This paper presents a computer vision guided peg hole insertion task conducted by a robot. We mounted two cameras: one capturing top view and the other capturing side view to calibrate the three dimensional co-ordinates of the center position of the peg and hole found in the image to the actual world co-ordinates so that the robot can grab and insert the peg into the hole automatically. We exploit normalized cross correlation based template matching and distortion model (grid) calibration algorithm for our experiment. We exploit a linear equation for the linear and rotational displacement of the arm and gripper of the robot respectively for computing the pulse required for the encoder. We utilize gantry robot to conduct the experiment. The implementation was carried out in LabVIEW environment. We achieved significant amount of accuracy with an experimental error of 5% for template matching and ±2.5 mm for calibration algorithm.