The Ship target recognition in complex marine environments faces challenges such as high noise interference and dynamic conditions, leading to frequent misclassifications and detection errors. This paper introduces an innovative method combining attention mechanisms with convolutional neural networks (CNNs) to enhance feature extraction and model robustness. By integrating multi-resolution analysis and feature fusion techniques, our proposed method prioritizes advantageous feature weights, improving both accuracy and reliability of ship positioning and model classification tasks. Experimental results on a specialized dataset demonstrate significant advancements over traditional recognition methods, validating the effectiveness of our approach in maritime surveillance applications.