Crack Detection and Localization with Real-Time Path Tracking using Stereo Vision for Autonomous Underwater Welding

Abstract

Oil and gas organizations across the globe rely heavily on underwater pipelines and water transportation systems. Due to exposure of pipe and ship surfaces to hostile environments like turbulence, water currents, and underwater vegetation, it is not uncommon to observe cracks on exposed material. Thus, regular inspection and maintenance are imperative. Such tasks need to be extremely precise which render the task to be time-consuming and the hostility of the environment renders the task to be dangerous and expensive. This calls for a robotic solution to the problem. In this paper, we propose a 7 DOF autonomous underwater welding arm along with autonomous crack detection, localization, and welding of the crack. We propose a real-time crack identification and segmentation technique using computer vision algorithms to facilitate real-time path generation for the arm, trajectory following, and motion planning. This methodology has been tested by mounting the designed arm on an AUV in the UnderWater Simulator and moving it using the MoveIt! package. The paper also implements a closed-loop control mechanism with an adaptive motion planner using pose tracking to overcome external disturbances during welding operations. The results indicate that the path tracking problem can be solved using computer vision algorithms in the absence of dynamic models.