Robotic Assembly Using 3D and 2D Computer Vision

Abstract

The content of this thesis concerns the development and evaluation of a robotic cell used for automated assembly. The automated assembly is made possible by a combination of an eye-inhand 2D camera and a stationary 3D camera used to automatically detect objects. Computer vision, kinematics and programming is the main topics of the thesis. Possible approaches to object detection has been investigated and evaluated in terms of performance. The kinematic relation between the cameras in the robotic cell and robotic manipulator movements has been described. A functioning solution has been implemented in the robotic cell at the Department of Production and Quality Engineering laboratory.

Theory with significant importance to the developed solution is presented. The methods used to achieve each part of the solution is anchored in theory and presented with the decisions and guidelines made throughout the project work in order to achieve the final solution.

Each part of the system is presented with associated results. The combination of these results yields a solution which proves that the methods developed to achieve automated assembly works as intended. Limitations, challenges and future possibilities and improvements for the solution is then discussed.

The results from the experiments presented in this thesis demonstrates the performance of the developed system. The system fulfills the specifications defined in the problem description and is functioning as intended considering the instrumentation used.