基于深度学习的海上舰船目标多源特征融合识别

Ship Targets Recognition Based on Deep Learning Multi-source Feature Fusion

针对海上舰船雷达辐射源特征和高分辨率距离像独立使用难以有效进行目标识别的问题,提出多种基于深度学习的多源特征融合目标识别方法。通过构造深度序贯融合模型、深度分支融合模型、深度卷积融合模型和深度循环融合模型四种框架,实现舰船目标的多辐射源特征与高分辨率距离像深度特征的自动提取与融合识别,完成目标分类。利用不同长度、不同类别和不同信噪比的仿真数据集对四种融合模型进行训练和测试,结果表明深度分支融合模型和深度循环特征融合模型具有较高的应用潜力和研究价值。

Because it’s difficult to recognize the ship targets using the feature of the radiation sources or the High Range Resolution Profiles( HRRP) individually,we propose four multi-source feature fusion methods based on deep learning to fuse the feature of radiation sources and HRRP and recognize the ship targets. The methods include the deep sequential feature fusion model,the deep branch feature fusion model,the deep convolution feature fusion model and the deep recurrent feature fusion model,which fuse the feature of multiple radiation sources and HRRP automatically and deeply. In order to test the performance of the four models,we construct a series of dataset using different target length,different numbers of categories and different signal-to-noise ratios. The results indicate that the deep branch feature fusion model and the deep recurrent feature fusion model have higher application potential and more practical research value.