With the growth of global liquefied natural gas (LNG) demand, automation technology has become a key trend to improve the efficiency and safety of LNG handling. In this study, a novel automatic docking system is proposed which adopts a staged docking strategy based on a monocular camera for positioning and combines ultrasonic sensors to achieve multi-stage optimization in the fine docking stage. In the coarse docking stage, the system acquires flange image data through the monocular camera, calculates 3D coordinates based on geometric feature extraction and coordinate transformation, and completes the preliminary target localization and fast approach; in the fine docking stage, the ultrasonic sensor is used to measure the multidirectional distance deviation, and the fusion of the monocular data is used to make dynamic adjustments to achieve high-precision alignment and localization. Simulation and experimental verification show that the system has good robustness in complex environments, such as wind and waves, and can achieve docking accuracy within 3 mm, which is better than the traditional manual docking method. This study provides a practical solution for automated docking of LNG loading arms, which can significantly improve the efficiency and safety of LNG loading and unloading operations.

Recognition of flanges on LNG (“Liquefied Natural Gas”) vessels and distance determination are key prerequisites for automated docking of an unloading arm, and in complex environments, flange detection faces challenges such as low accuracy and large distance measurement errors. Therefore, an improved lightweight high-precision approach based on YOLOv8s that integrates the C2f_Ghost module, a CBAM, and Concat_BiFPN was proposed. In addition, a monocular ranging algorithm based on pixel width and coordinate transformation was introduced to estimate the 3D coordinates of a flange. Specifically, the original Bottleneck in the C2f module was replaced by the Ghost module, which combines dynamic convolution and dynamic depth-separable convolution to enhance feature representation while reducing model complexity. A CBAM was introduced in the middle layer of the backbone module to improve the model’s focus on key features with minimal parameter increases, and Concat_BiFPN was used in the neck to facilitate cross-scale feature fusion. To ensure the reproducibility of the experiments, this study primarily employed a fixed random seed (0), and experimental validation analysis was conducted on a flange dataset. The results show that for the improved model, mAP@0.5 reached 97.5% and mAP@0.5:0.95 reached 82.3% with a parameter size of 9.34 M, representing improvements in accuracy of 0.6% and 13.4%, respectively, and a reduction in the parameter size of 16.2% compared with the original YOLOv8s. The average ranging errors of the X-axis, Y-axis, and Z-axis were 2.43%, 2.77%, and 0.71%, respectively. Therefore, the combination of the two algorithms significantly improves the detection and ranging accuracy of flanges in complex environments.