In the task of classifying images of cracks in underwater dams, symmetry serves as a crucial geometric feature that aids in distinguishing cracks from other structural elements. Nevertheless, the asymmetry in the distribution of positive and negative samples within the underwater dam crack image dataset results in a long-tail problem. This asymmetry, coupled with the subtle nature of crack features, leads to inadequate feature extraction by existing convolutional neural networks, thereby reducing classification accuracy. To address these issues, this paper improves VanillaNet. First, the Seesaw Loss loss function is introduced to tackle the long-tail problem in classifying underwater dam crack images, enhancing the model’s ability to recognize tail categories. Second, the Adaptive Frequency Filtering Token Mixer (AFF Token Mixer) is implemented to improve the model’s capability to capture crack image features and enhance classification accuracy. Finally, label smoothing is applied to prevent overfitting to the training data and improve the model’s generalization performance. The experimental results demonstrate that the proposed improvements significantly enhance the model’s classification accuracy for underwater dam crack images. The optimized algorithm achieves superior average accuracy in classifying underwater dam crack images, showing improvements of 1.29% and 0.64% over the relatively more accurate models ConvNeXtV2 and RepVGG, respectively. Compared to VanillaNet, the proposed algorithm increases average accuracy by 2.66%. The improved model also achieves higher accuracy compared to the pre-improved model and other mainstream networks.

As critical infrastructure, hydropower dams require efficient and accurate detection of underwater structural surface defects to ensure their safety. This paper presents the design and implementation of a robotic control system specifically developed for underwater dam inspection in hydropower stations, aiming to enhance the robot’s operational capability under harsh hydraulic conditions. The study includes the hardware design of the control system and the development of a surface human–machine interface unit. At the software level, a modular architecture is adopted to ensure real-time performance and reliability. The solution employs a hierarchical architecture comprising hardware sensing, real-time interaction protocols, and an adaptive controller, and the integrated algorithm combining a fixed-time disturbance observer with adaptive super-twisting controller compensates for complex hydrodynamic forces. To validate the system’s effectiveness, field tests were conducted at the Baihetan Hydropower Station. Experimental results demonstrate that the proposed control system enables stable and precise dam inspection, with standard deviations of multi-degree-of-freedom automatic control below 0.5 and hovering control below 0.1. These findings confirm the system’s feasibility and superiority in performing high-precision, high-stability inspection tasks in complex underwater environments of real hydropower dams. The developed system provides reliable technical support for intelligent underwater dam inspection and holds significant practical value for improving the safety and maintenance of major hydraulic infrastructure.

Hydropower stations and dams play a crucial role in water management, ecology, and energy. To meet the requirements of underwater dam defect detection, this study develops a streamlined underwater vehicle design and operational framework inspired by bionic principles. A parametric modeling approach was employed to propose the vehicle’s streamlined configuration. Using CFD simulations, hydrodynamic coefficients were calculated and validated through towing experiments in a pool. The hydrodynamic stability of the vehicle was assessed and verified through these analyses. Additionally, various configurations were generated using a free deformation method. An optimization function was established with resistance and stability as the objectives, and the optimal result was derived based on the function’s calculation outcomes. The study designed a high-metacentric underwater vehicle, inspired by the seahorse’s shape, and introduced a novel stability evaluation method. Simulations were conducted to analyze the vehicle’s variable attack angle, drift angle, pitching, and rotational motion at a forward three-throttle speed. The results demonstrate that the vehicle achieves static stability in both the horizontal and vertical planes, as well as dynamic stability in the vertical plane, but exhibits limited dynamic stability in the horizontal plane. After optimizing the original configuration, the forward resistance was reduced by 2.15%, while the horizontal plane dynamic stability criterion C_H was improved by 35.29%.

In adverse foggy weather conditions, images captured are adversely affected by natural environmental factors, resulting in reduced image contrast and diminished visibility. Traditional image dehazing methods typically rely on prior knowledge, but their efficacy diminishes in practical, complex environments. Deep learning methods have shown promise in single-image dehazing tasks, but often struggle to fully leverage depth and edge information, leading to blurred edges and incomplete dehazing effects. To address these challenges, this paper proposes a deep-guided bilateral grid feature fusion dehazing network. This network extracts depth information through a dedicated module, derives bilateral grid features via Unet, employs depth information to guide the sampling of bilateral grid features, reconstructs features using a dedicated module, and finally estimates dehazed images through two layers of convolutional layers and residual connections with the original images. The experimental results demonstrate the effectiveness of the proposed method on public datasets, successfully removing fog while preserving image details.

Sonar images possess characteristics such as low resolution, high noise, and blurred edges. Utilizing CNNs would lead to problems such as inadequate target recognition accuracy. Moreover, due to their larger sizes and higher computational requirements, existing CNNs face deployment issues in embedded devices. Therefore, we propose a sonar image recognition algorithm optimized for the lightweight algorithm, MobileViT, by analyzing the features of sonar images. Firstly, the MobileViT block is modified by adding and redesigning the jump connection layer to capture more important features of sonar images. Secondly, the original 1 × 1 convolution is replaced with the redesigned multi-scale convolution Res2Net in the MV2 module to enhance the ability of the algorithm to learn global and local features. Finally, the IB loss is applied to address the imbalance of sample categories in the sonar dataset, assigning different weights to the samples to improve the performance of the network. The experimental results show that several proposed improvements have improved the accuracy of sonar image recognition to varying degrees. At the same time, the proposed algorithm is lightweight and can be deploy on embedded devices.

This paper presents a positioning control method for underactuated unmanned surface vehicles (USVs) subject to environmental disturbances and wave-contaminated measurements. In underactuated dynamic positioning, surge motion and yaw motion can be directly regulated by the propulsion system, whereas sway motion cannot be directly controlled by an independent lateral thrust. Therefore, the lateral environmental-force component is utilized to induce the vehicle’s dynamic response to sway. To achieve this objective, a position-keeping guidance system taking into account the desired heading and the lateral positioning error is introduced in this paper. With this guidance mechanism, the lateral environmental-force component can drive the USV to reduce the cross-track error, thereby enabling underactuated positioning. A rotated coordinate system is established around the desired position, and the positioning error is decomposed into along-track and cross-track components. To improve the transient response, an error-rate feedback term is introduced into the rotated-angle update law for yaw-heading guidance design, which enhances the damping of the cross-track dynamics. Meanwhile, a wave-filtering observer is designed to make low-frequency position and velocity estimates for feedback control. Simulation results under multiple operating conditions show that the proposed observer reduces the amplitude and high-frequency variation of the control signals compared with the existing wave-filtering observer, and the proposed positioning control method achieves smaller positioning errors than the existing nonlinear positioning control (NPC). The comparative results also indicate that the proposed method is suitable for position keeping under constant or slowly varying environmental loads, moderate model uncertainty, and wave-contaminated measurements, whereas rapidly varying load directions may degrade the positioning accuracy.