Industrial robots are the key element in today's automation. One of the most common ro-botic applications in the industrial sector is robot welding. Robot-based automation increases the efficiency of welding processes and enables manufacturing more parts in less time, while ensuring welding quality in the end. To get a highly automated welding process, it is necessary to control and monitor the process. Because of numerous errors such as defects that could happen during the process, thermal deformation of the workpiece, positioning, and manu-facturing errors, the welding robots need to be positioned and adjusted in real-time to reduce these errors. This requires monitoring and seam tracking techniques. Welding process is complex, and even though in recent years industrial robotic welding has evolved rapidly, there has always been some complications, obstacles, and challenges along the way to im-prove automated welding process.Siemens Energy AB is manufacturer of world-class, producing low environmental impactsteam and gas turbines for industrial use with high efficiency and low emission levels. Sie-mens Energy AB also has a manufacturing unit in Trollhättan with specialty in welding, pressing, shaping, laser machining stainless, nickel alloy, and heat resistant materials. This unit mainly manufactures combustion chambers for the gas turbines.At Trollhättan site, there is a need to automate the welding station and there have been some difficulties with the vision system, which prevents the welding process to detect the groove's edges and work as expected. This master thesis is an experimental study that in the first step aims to find and assess the problems with the old vision system at the company, and later propose new approaches to solve them by implementing and utilizing optical sensors to obtain target surface data required to realize an automated arc welding process. To achieve this, a vision system is proposed based on a laser line and camera, and the acquired images by the camera are processed using different methods and algorithms. It is shown that the proposed vision system can detect the groove's edges and extract feature points on the work piece with no difficulties.