Navigating the complexities of ship movements and the wide array of vessel types poses a considerable obstacle for the automatic extraction and classification of information from ship trajectory and motion data. This paper proposes a sophisticated hybrid deep learning approach that efficiently extracts and integrates key features from ship trajectory images and motion data sequences. By leveraging the combined strengths of Convolutional Neural Networks (CNN), Bidirectional Gated Recurrent Units (BiGRU), and Transformer encoders, this model offers a novel solution for the precise classification of ship types. Rigorously tested on a real-world Automatic Identification System (AIS) dataset, the model demonstrates a notable improvement in classification accuracy over existing methods. This research not only provides an effective tool for automated ship classification in maritime traffic management but also paves the way for the application of advanced multimodal data processing and deep learning techniques in complex classification scenarios.