Vision-based Object’s Shape Determination for Robot Alignment

Abstract

This study provides vision-based system solutions for a peg-in-hole problem faced by a fork lift like robot used to transport copper wire spools from a rack, in which the spools are arranged side by side to a specified place. The copper wire spool (a cylindrical object on which the copper wire is wound and have a rim at each end) is held by 3 cylindrical shafts; one of the shafts is inserted through the center hole of the spool and another two shafts is held at the bottom of the spool. The aim of the development of vision-based system is to enable the robot to pick up the spool autonomously. To enable the center cylindrical shaft to be inserted nicely through the center hole of the spool, the center point of the spool must be on the center line of the camera Field of View (FOV). The problem to be solved in this study is how to determine that the center point is overlapped with the center line of the camera FOV. Firstly, a circle with the same radius of the spool’s rim was created at the center of the camera frame on screen, and then the spool’s front rim was tracked until it is overlapped with the circle on the screen image to ensure it is on the line of the camera FOV. However, the scope of this paper is limited to copper wire spool detection, and the confirmation of the front rim overlapping conditions is based on real time video processing. The proposed system uses Circular Hough Transform (CHT), binarization, morphology and edge detection of the sampled images from real-time video recording. A Logitech Webcam C270, which has an autofocus camera and HD view with lower price is used. By integrating the Logitech webcam for windows with MATLAB R2016a, all computations, programming and processing of this project are done using the MATLAB. Several experiments had been carried out and from the result obtained, the system is able to track the spool and determine the correct position of the robot to pick up the spool.