A ship roll prediction scheme is proposed using an adaptive sliding data window (SDW), which is designed to represent time-varying nonlinear dynamics of ship roll motion. The adjustment of SDW is realized by developing an improved fuzzy Gath-Geva (IFGG) segmentation approach, which detects the changes of system dynamics and thereby automatically adapting the scale of SDW. By virtue of the learning scheme with an adaptive SDW, the variable-structure radial basis function network is constructed sequentially to online predict ship roll dynamics. Experimental studies on online ship roll prediction are conducted on measured data from YuKun's full-scale sea trial. Results demonstrate the remarkable predictive accuracy of the proposed ship roll prediction model as well as the effectiveness of the IFGG-based SDW in terms of representing time-varying dynamics.

Due to the recent rapid growth of ocean oil development and transportation, the offshore oil spill risk accident probability has increased unevenly. The marine oil spill poses a great threat to the development of coastal cities. Therefore, effective and reliable technologies must be used to monitor oil spills to minimize disaster losses. Based on YOLO deep learning network, an automatic oil spill detection method was proposed. The experimental data preprocessing operations include noise reduction, gray adjustment, and local contrast enhancement. Then, real and synthetically generated marine radar oil spill images were used to make slice samples for training the model in the YOLOv5 network. The detection model can identify the effective oil spill monitoring region. Finally, an adaptive threshold was applied to extract the oil slicks in the effective oil spill monitoring regions. The YOLOv5 detection model generated had the advantage of high efficiency compared with existing methods. The offshore oil spill detection method proposed can support real-time and effective data for routine patrol inspection and accident emergency response.

Tide variations are affected not only by periodic movement of celestial bodies but also by time-varying interference from the external environment. To improve the accuracy of tide prediction, a modular tide level prediction model (HA-NARX) is proposed. This model divides tide data into two parts: astronomical tide data affected by celestial tide-generating forces and nonastronomical tide data affected by various environmental factors. Final tide prediction results are obtained using a nonlinear autoregressive exogenous model (NARX) neural network combined with harmonic analysis (HA) data. To verify the feasibility of the model, tide data under different climatic and geographical conditions are used to simulate the prediction of tide levels, and the results are compared with those of traditional HA, the genetic algorithm-back propagation (GA-BP) neural network and the wavelet neural network (WNN). The results show that the greater the influence of meteorological factors on tides, the more obvious is the improvement in accuracy and stability of HA-NARX prediction results compared to traditional models, with the highest prediction accuracy improvement of 234%. The proposed model not only has a simple structure but can also effectively improve the stability and accuracy of tide prediction.