Weld seam detection based on visual saliency for autonomous welding robots

Abstract

Nowadays autonomous welding robots are gaining importance. Weld seam detection is a key technology in robotic welding which is usually performed using a visual system. In most existing vision-based approaches, traditional image processings are used to obtain weld seams. Generally, these approaches are sensitive to their surrounding environment, especially illumination. Therefore, we address the problem by introducing the visual attention mechanism of the primate. Firstly, image preprocessing is executed to block visual interferences. Secondly, a visual saliency model based on local contrast is proposed to emphasize weld seam candidates. Finally, some basic image processings are performed to extract the desired weld seam. In order to validate the proposed approach, experiments are carried out in different cases: two types of joints (butt and fillet joint) and three types of shapes (straight line, zigzag and curve). The results demonstrate that this method is effective and robust, and is useful for autonomous welding robots.