<p>The statistical data on maritime accidents indicate that human factors are the primary cause of such incidents. To improve crew members’ ability to handle emergencies, this study explores the application of brain–computer interface technology in crew emotion recognition. Utilizing virtual reality technology, three maritime emergency scenarios—fire, severe sea conditions, and lifeboat deployment—are simulated to collect the electroencephalogram (EEG) signals of participants. The signals are then preprocessed. Four advanced deep learning models in the field of emotion recognition are constructed to classify and identify EEG signals. A novel crew fear emotion recognition method, EEG-MCnet, suitable for emergency scenarios, is proposed. Multiple evaluation metrics, including accuracy, precision, F1-score, recall, and Kappa coefficient, are used to comprehensively compare the performance of the models. Experimental results show that EEG-MCnet outperforms four existing methods in the three-class crew emotion recognition task, demonstrating excellent performance across all evaluation metrics with an accuracy of 86.30%. This method effectively extracts key features from EEG signals, accurately identifies the fear emotion of crew members in emergency situations, and provides targeted feedback and training suggestions, thereby enhancing the crew’s emergency response capabilities in real-world maritime operations.</p>