Research on Image-Based 3D Reconstruction Method in Underwater Welding Scenes

Abstract

The working environment of underwater structures is complex and often requires underwater welding repair. This paper proposed an underwater image-based 3D reconstruction method for underwater dry welding robots, which can be applied to welding scenes in low-light underwater environments. Firstly, this paper used commonly used binocular and RGB-D cameras to conduct 3D reconstruction tests in air, wet, and underwater scenes, respectively. The conclusion was drawn that binocular cameras have higher accuracy in underwater scenes, while RGB-D cameras have higher accuracy in air and wet scenes. Secondly, an improved binocular image-based 3D reconstruction method was proposed and tested on the Haud public dataset. Finally, a new underwater binocular-RGB-D image-based 3D reconstruction method was proposed for an underwater welding scene. The method proposed in this paper can provide a prerequisite for the automation development of underwater welding robots.