Multicamera–IMU Calibration System for Underwater Welding Scenes

As artificial underwater welding is dangerous and inefficient, visual-based pose estimation and 3-D reconstruction can provide accurate weld position information, which can replace human eye observation and greatly improve the degree of welding automation. The calibration accuracy of camera intrinsic parameters and extrinsic parameters with an inertial measurement unit (IMU) directly affects the accuracy of underwater pose estimation. Under the influence of complex underwater environments, the accuracy of underwater calibration is often insufficient, and the efficiency of multisensor calibration is unsatisfactory. To solve the above problems, an underwater multicamera–IMU calibration system is proposed. First, to solve the problem of unstable checkerboard corner recognition during underwater intrinsic parameter calibration, a prediction-detection method is proposed. Second, an intrinsic–extrinsic parameters joint optimization method is proposed to improve the accuracy and efficiency of the calibration. Then, a system for multicamera–IMU calibration was established, further improving the efficiency. Finally, extensive experiments based on public datasets and welding scene datasets were used to verify the feasibility and effectiveness of the proposed method.