Image Processing for Automated Welding Robot: Reducing Interference due to Fume in Camera Lenses

Welding is crucial for several industrial sectors. The problems associated with this laborious work and the worker's health are known for a long time. However, the welding is still highly dependent of a skilled human workforce. Researchers presented solutions to automate this process, many of them using robots equipped with cameras. Nevertheless, the fumes produced by the arc welding process adhere to the camera lens hinder the robot's perception. The fume is composed of by-products from melting metal becoming suspended in the air and adhering to surrounding surfaces. Thus, one of the main issues to be solved in the robotized welding with cameras is to deal with the fume adhered the camera lenses. In this paper, we propose a method to minimize the interference of the fumes present in the camera lens based on a fusion of image processing methods. The results shows the method is able to improve the groove detection allowing a better welding.