Qinyou Hu

In the context of global efforts to promote energy conservation and emission reduction, geopolitical conflicts have intensified the challenges of mitigating marine climate change, posing increasingly severe economic and climatic pressures on the shipping industry worldwide. Research on multi-objective route optimization is of great significance for addressing climate challenges and enhancing economic efficiencies. This field focuses on constructing multi-objective optimization models that aim to reduce voyage time, fuel consumption, navigational risks, and carbon emissions and solving them using various algorithms. However, determining the optimal route and sailing speed under complex and variable meteorological conditions remains a significant challenge owing to the presence of numerous unevenly distributed feasible solutions within a vast solution space, making it difficult for traditional intelligent algorithms to effectively explore this space. To address this issue, this study proposes a hybrid algorithm named NRRT by integrating the Rapidly exploring Random Tree (RRT) algorithm with the Non-dominated Sorting Genetic Algorithm III (NSGA-III). By improving the sampling logic of the RRT algorithm and combining the vessel’s voluntary speed loss with the sampling step size, the algorithm efficiently explored the feasible route set, enhancing the quality and diversity of the solutions. Subsequently, the NSGA-III algorithm treats sailing speed and heading as direct decision variables to perform multi-objective optimization on the explored routes and generate Pareto-optimal solutions. The optimization results demonstrate that the proposed method excels at generating route plans that effectively reduce costs, minimize emissions, and mitigate risks compared with the 3D Dijkstra algorithm and the improved NSGA-III algorithm.

The occurrence of sea fog leads to a sharp reduction in visibility, posing threats to maritime navigation and port operations. Existing visibility classification methods are predominantly developed for land scenarios, where the categorization of fog conditions remains coarse and is difficult to directly apply to the complex maritime environment. The scarcity of sea fog visibility datasets further exacerbates this challenge. To address these issues, this study proposes a fine-grained sea fog visibility classification network, M-ConvNeXt, designed for multi-level and real-time assessment in support of maritime safety. First, a frequency-domain perturbation strategy was developed to enhance sample diversity and improve model robustness. Then, an adaptive inverted residual block with channel attention was introduced to better extract fine-grained features of sea fog images across different categories. In addition, a maritime visibility image dataset covering six levels was constructed. Experimental results demonstrate that M-ConvNeXt achieves high accuracy and delivers robust performance in fine-grained sea-fog visibility classification, with an average accuracy of 94.59%. This represents improvements of 8.06% over the baseline ConvNeXt-T, 6.83%, 5.97%, and 5.96% over CNN-based ResNeXt, ResNet50 and EfficientNet-B0 respectively, and 2.56% over Swin-Transformer.

In the context of vessel monitoring, the accuracy of vessel speed measurements is contingent on the availability of AIS data. However, the absence, failure, or signal congestion of AIS devices may lead to delays and inaccuracies in the speed information. To address this challenge, this paper proposes a vessel speed detection method based on target detection and tracking to acquire vessel speed in real time. The proposed methodology involves the establishment of a mapping relationship between image coordinates and four real-world coordinates, ensuring precise conversion from pixel velocity to physical velocity. Subsequently, a frame difference method combined with a multi-frame averaging strategy calculates the vessel speed. Furthermore, an advanced M-YOLOv11 detection model is introduced to enhance the detection performance in different vessel shapes and complex environments, thus ensuring the accuracy of speed information is further improved. The experimental results demonstrate that M-YOLOv11 exhibits a significant performance enhancement, with a 13.95% improvement in the average precision metric over the baseline model. Over 60% of the measured vessel speed measurement errors are less than 0.5 knots, with an overall average error below 0.45 knots. These findings substantiate the efficacy and superiority of the proposed method in practical applications.

In this study, we developed an interpretable machine learning (ML) framework to predict marine pollution and economic losses from accident risk factors. A triple-feature selection process identified key predictors, followed by a comparative analysis of eight ML models. Random forest outperformed other models in forecasting environmental and property damage. The interpretable model was established based on the SHAP value framework, which revealed that onboard personnel count, vessel dimensions (length), and accident/ship types account for the risk factors with the most severe consequences, with environmental conditions like wind speed and air temperature contributing secondary effects. The methodology enables two critical applications: (1) environmental agencies can proactively assess accident impact through the identified risk triggers, optimizing emergency response planning, and (2) insurance providers gain data-driven risk evaluation metrics to refine premium calculations. By quantifying how human/technical factors, including crew members and vessel specifications, dominate over natural variables in accident effects, this data-driven approach provides actionable insights for maritime safety management and financial risk mitigation, achieving high prediction accuracy while maintaining model transparency through Shapley value explanations.

The ice conditions in Arctic waters are complex and variable, requiring ships to dynamically adjust their routes to ensure safe and efficient navigation. Traditional dynamic path planning algorithms struggle to address the extensive variability of Arctic ice conditions. To tackle this issue, this paper improves the D* Lite algorithm by leveraging the gradual and convergent nature of Arctic ice condition changes. The original algorithm’s local update and path extraction rules are modified to prevent chain updates triggered by minor localized changes, thereby reducing the frequency of updates in non-critical areas. By simulating dynamic route planning for ships in Arctic waters during both the freezing and melting periods, the improved D* Lite algorithm was compared with the original D* Lite algorithm and a global update algorithm in terms of voyage distance, risk coefficient, planning time, and the number of node updates. The computational results demonstrate that the improved D* Lite algorithm achieves planning results very similar to those of the original D* Lite algorithm and the global update algorithm at the lowest update cost, significantly enhancing the safety and efficiency of dynamic route planning for ships in Arctic waters.

Ship fuel consumption plays a crucial role not only in understanding ships’ energy efficiency but also in gaining insights into their emissions. However, enhancing the accuracy of these predictions poses significant challenges due to data limitations and the methods employed. Due to these factors, such as data variability and equipment characteristics, ship fuel consumption exhibits certain fluctuations under specific conditions. Previous fuel consumption prediction methods primarily generate a single specific value, making it difficult to capture the volatility of and variability in fuel consumption. To overcome this limitation, this paper proposes a novel method that integrates Gaussian process prediction with quantile regression theory to perform interval predictions of ship fuel consumption, providing a range of possible outcomes. Through comparative analyses with traditional methods, the possibility of using the method is verified and its results are validated. The results indicate the following: (1) at a 95% confidence level, the proposed method achieves a prediction interval coverage probability of 0.98 and a prediction interval normalized average width of 0.123, which are significantly better than those of the existing backpropagation neural network (BPNN) and gradient boosting decision tree (GBDT) quantile regression models; (2) the prediction accuracy of the proposed method is 92% for point forecasts; and (3) the proposed method is applicable to main datasets, including both noon report and sensor datasets. These findings provide valuable insights into interval predictions of ship fuel consumption and highlight their potential applications in related fields, emphasizing the importance of accurate interval predictions in intelligent energy efficiency optimization.

With the development and application of artificial intelligence (AI) in the shipping industry, using AI to replace traditional draft survey methods in bulk carriers can significantly reduce manpower, lower the risks associated with visual observations, improve measurement accuracy, and minimize the impact of human subjective factors. Ultimately, the integration of software and hardware technologies will replace human visual observations with automated draft measurement calculations. A similar anti-fluctuation device described in this article has been used in ship draft observation based on AI-assisted proving, which can ease the fluctuation of the wave inside the pipe. Observers can directly read the water surface inside the pipe and compare it to the ship’s draft mark to obtain the final draft, effectively improving draft observation accuracy. However, some surveyors refuse to accept the readings obtained from this device, citing a lack of theoretical basis or the absence of accreditation from relevant technical authorities, leading to the rejection of results. To address these issues, this paper integrates wave energy attenuation theory with PaddlePaddle-OCR recognition to further validate the anti-fluctuation device for accurate ship draft observation. The experimental results are as follows: first, the pipe effectively suppresses the amplitude of external water surface fluctuations by 75%, explaining the fundamental theory that wave heights within the anti-fluctuation device are consistent with external swell heights. When taking a draft measurement, the system dynamically adjusts the position of the main tube in response to the ship’s movements, maintaining the stability of the measurement section and significantly reducing the difficulty of observations. Due to the reduction in fluctuation amplitude, there is a noticeable improvement in observation accuracy.

Accurate geolocation of maritime objects in satellite imagery is challenging due to geometric distortions, atmospheric conditions, and sensor inaccuracies in low-earth orbit satellites. This study presents a novel AIS data-guided geolocation correction method that integrates real-time Automatic Identification System (AIS) data with satellite imagery to rectify geolocation errors. The approach utilizes the GeoAISNet neural network, which enhances positional accuracy without relying on Ground Control Points (GCPs). By incorporating a modified YOLOv8 architecture with orientation parameters and the Convolutional Block Attention Module (CBAM), detection performance improved significantly, achieving precision, recall, and F1 scores of 91.82%, 89.56%, and 90.67%, respectively. Ablation studies demonstrated the crucial impact of feature integration and attention mechanisms. Results indicate a mean Average Precision (mAP) of 89%, with General Cargo Ships achieving 99.9% AP50. Localization accuracy saw a notable improvement, with RMSE reduced from 12 meters to 3 meters, and layer normalization further enhanced stability, increasing precision, recall, and F1 scores to 94.23%, 92.67%, and 93.44%, respectively. The use of differential AIS data decreased maximum positional errors by 30%, achieving errors around 2 meters. Computational efficiency was also enhanced, with processing time reduced from 2 seconds to 0.5 seconds per image. This method effectively addresses oil spills and non-AIS vessel detection, expanding maritime surveillance capabilities. The global training dataset, validated with data from the South China Sea, ensures the method's applicability across diverse conditions.