Machine vision camera calibration and robot communication

Abstract: This thesis is a part of a larger project included in the European project, AFFIX. The reason for the project is to try to develop a new method to assemble an aircraft engine part so that the weight and manufacturing costs are reduced. The proposal is to weld sheet metal parts instead of using cast parts. A machine vision system is suggested to be used in order to detect the joints for the weld assembly operation of the sheet metal. The final system aims to locate a hidden curve on an object. The coordinates for the curve are calculated by the machine vision system and sent to a robot. The robot should create and follow a path by using the coordinates. The accuracy for locating the curve to perform an approved weld joint must be within +/- 0.5 mm. This report investigates the accuracy of the camera calibration and the positioning of the robot. It also brushes the importance of good lightning when obtaining images for a vision system and the development for a robot program that receives these coordinates and transform them into robot movements are included. The camera calibration is done in a toolbox for MatLab and it extracts the intrinsic camera parameters such as the distance between the centre of the lens and the optical detector in the camera: f, lens distortion parameters and principle point. It also returns the location of the camera and orientation at each obtained image during the calibration, the extrinsic parameters. The intrinsic parameters are used when translating between image coordinates and camera coordinates and the extrinsic parameters are used when translating between camera coordinates and world coordinates. The results of this project are a transformation matrix that translates the robots position into the cameras position. It also contains a robot program that can receive a large number of coordinates, store them and create a path to move along for the weld application.