Liming Li

Subsea Christmas trees are often deployed in turbid coastal waters or seabed environments. During manipulator operations on Christmas tree panels, conventional optical servoing is severely limited by rapid electromagnetic attenuation and strong scattering from suspended particles, resulting in reduced visibility. Forward-looking sonar (FLS) provides stable imaging, but its unique imaging geometry and low resolution make direct 6D pose estimation challenging. To address this issue, this paper proposes a 6D object pose estimation method for FLS images, in which conventional optical control-point-based pose estimation is restructured to resolve the mismatch between optical-centric network assumptions and acoustic imaging characteristics, and is further integrated with acoustic projection-based pose inversion. First, to address the limited diversity of target appearances and the scarcity of training data, we construct an FLS imaging model based on primary truncation for image simulation, providing data for model pretraining. Second, a multi-task acoustic control-point detection network, Acoustic-Yolo6D, is designed to mitigate localization degradation caused by heavy speckle noise, low boundary contrast, and resolution variations associated with polar-coordinate imaging, through heatmap regression, auxiliary object segmentation, and explicit range-bearing positional encoding. An Acoustic-n-Point (AnP) model is then used to recover the target 6D pose. Finally, simulation and water-tank experiments on the socket target verify the feasibility and robustness of the proposed method under limited-data conditions. The method achieves a 3.1 cm mean translation error, a 10.88° mean orientation error, and 52 FPS in real underwater acoustic environments.

Side‐scan sonar is essential for underwater observation and seabed target detection, yet traditional single‐frequency systems must compromise between imaging resolution and detection range. For object detection tasks, high‐frequency sonar yields finer details contributed by stronger scattering and higher contrast but suffers from limited coverage due to heavier attenuation, whereas low‐frequency sonar covers wider areas yet offers poorer small‐target visibility. To overcome these trade‐offs and the challenges of strong speckle noise, target–shadow coupling and resolution variation, this study proposes a dual‐frequency detection model named D ² FNet (dual‐domain fusion network). D ² FNet integrates three key modules: (1) a union domain attention (UDA) module for preliminary dual‐frequency fusion via ResNet‐50 and a transformer encoder; (2) a cross‐domain attention (CDA) module for enhanced feature interaction across frequency domains; and (3) a target–shadow pairing (TSP) module that embeds sonar imaging priors through local window attention to suppress false alarms and improve localisation confidence. Based on a newly constructed dual‐frequency side‐scan sonar dataset containing over 9000 paired images from sea trials, D ² FNet significantly outperforms single‐frequency and baseline fusion models in mAP metrics, demonstrating its effectiveness for high‐precision underwater target detection.