Synthetic aperture radar (SAR), which can work normally under various meteorological conditions, has been widely researched and applied in marine vessel target monitoring and identification. Among the many research topics, due to the inconsistency of ship scale in SAR images, susceptibility to sea and land noise and clutter interference, resulting in a low detection rate of near-shore ship targets and inaccurate edge delineation of densely lined ships, a target detection algorithm based on deep convolutional neural network is proposed. The algorithm employs the channel-space grouping attention mechanism during feature extraction to enhance features by utilizing global positional and edge information associated with instances. The feature mobility fusion module is employed to merge features of various scales, bolster the interconnection among these features, and enhance multiscale ship target detection capabilities. The decoupled head is employed for ship target localization, while the angle-weighted intersection over union is used to mitigate regression errors. The experimental results show that the precision (P) achieved on HRSID and SSDD datasets reaches 94.81% and 99.01%, respectively, exceeding the control algorithm by more than 1.35% and 0.94%; the average precision (mAP) reaches 92.06% and 99.50%, respectively, exceeding the control algorithm by more than 2.32% and 2.51%; this indicates that the proposed algorithm has a good performance on SAR image ship detection and a strong generalization ability.

Sudden collapse accidents in mine roadways occur frequently, and post-disaster emergency rescue faces major challenges in terms of safety and efficiency. Therefore, efficient demolition equipment and intelligent prediction methods are urgently needed. Abrasive waterjet technology has considerable potential for complex disaster environments owing to its high efficiency, environmental friendliness, and cold-cutting characteristics. However, its cutting performance is affected by multiple coupled factors, including jet parameters, material properties, and environmental conditions. This makes accurate prediction difficult, especially under extreme or unseen operating conditions where available samples are limited. To address this problem, this study proposes a zero-shot learning-based multi-physics coupling prediction framework for the “jet–material–environment–effect” relationship. The framework is designed to predict abrasive waterjet cutting performance under unseen working conditions. First, a multi-factor cutting-performance dataset is constructed through a hierarchical experimental design. A generative adversarial network (GAN) is then introduced to expand the sample space and compensate for the discrete nature and limited distributional coverage of the experimental data. Second, a lightweight self-attention mechanism is employed to model high-dimensional input features globally, thereby improving the model’s ability to capture complex feature interactions. Finally, a joint loss function is designed to collaboratively optimize the generation and prediction processes. The experimental results show that the proposed model achieves a prediction accuracy of 98.3% on the test set, with a coefficient of determination R² of 0.967, outperforming WOA-SVM, BP neural network, EML, and Transformer models. The inference response time is approximately 3.2 s, indicating good engineering applicability. The results demonstrate that GAN effectively expands the sample space and improves model generalization, while the LightTransformer structure provides advantages in modeling high-dimensional coupled inputs. The proposed method can provide theoretical support and technical reference for intelligent demolition rescue and cutting-depth prediction under mine disaster conditions.