在复杂海域场景下如何综合利用舰船监测的多模态数据进行高效特征提取和特征融合, 以此来综合提升舰船识别精度仍存在巨大挑战。针对海域环境中舰船单一数据源识别准确率问题, 提出一种有效的多模态数据特征提取和特征融合的舰船识别算法, 然后基于深度残差网络模型进行特征融合以提升舰船识别准确率。通过实验结果对比, 相比于其他算法基于多模态数据的舰船识别算法平均准确率提升约18%, 有效地提升了舰船识别准确率, 对相关船舶领域的研发工作具有借鉴意义。

There are huge challenges in how to comprehensively utilize multi-modal data from ship monitoring in complex sea area scenarios for efficient feature extraction and feature fusion to comprehensively improve ship identification accuracy. Aiming at the problem of ship identification accuracy from a single data source in the maritime environment, an effective ship identification algorithm for multi-modal data feature extraction and feature fusion is proposed, and then feature fusion is performed based on a deep residual network model to improve the accuracy of ship identification rate. Through comparison of experimental results, compared with other algorithms, the average accuracy of the ship identification algorithm based on multi-modal data is increased by about 18%, which effectively improves the accuracy of ship identification and has reference significance for research and development in related ship fields.