| 摘要: |
| 针对合成孔径雷达(Synthetic Aperture Radar,SAR)图像目标识别问题，提出了一种基于集成卷积神经网络(Convolutional Neural Network,CNN)的SAR图像目标识别方法。首先对原始数据集进行数据增强的预处理操作，以扩充训练样本；接着通过重采样的方法从训练样本中获取不同的训练子集，并在训练各基分类器时引入Dropout和Padding操作，有效增强了网络泛化能力；然后采用Adadelta算法与Nesterov动量法结合的思想来优化网络，提高了网络的收敛速度和识别精度；最后采用相对多数投票法对基分类器的分类结果进行集成。在MSTAR数据集上进行的实验结果表明，集成后的模型识别准确率达到99.30%，识别性能优于单个卷积神经网络，具有较强的泛化能力和较好的稳健性。 |
| 关键词: 雷达目标识别 合成孔径雷达（SAR） 卷积神经网络(CNN) Ada_Nesterov动量法 网络集成 |
| DOI：10.20079/j.issn.1001-893x.220427007 |
|
| 基金项目:国家自然科学基金资助项目(61871203) |
|
| A SAR Image Target Recognition Algorithm Based on Ensemble Convolutional Neural Network |
| LI Wanghua,ZHANG Zhenkai |
| ((Ocean College,Jiangsu University of Science and Technology,Zhenjiang 212100,China)) |
| Abstract: |
| For the problem of target recognition in synthetic aperture radar(SAR) image target recognition,a method of SAR image target recognition based on the ensemble convolutional neural network(CNN) is proposed.Firstly,the data-enhanced preprocessing operation of the original data set is performed to expand the training samples.Secondly,different training subsets are obtained from the training samples by resampling,and Dropout and Padding operations are introduced when training each base classifier,which effectively enhances the network generalization ability.Then the idea of combining the Adadelta algorithm and the Nesterov momentum algorithm is used to optimize the network,which improves the convergence speed and recognition accuracy of the network.Finally,the relative majority voting method is used to combine the classification results of base classifier.Experiment results on the MSTAR dataset show that the recognition accuracy of the ensemble model reaches 99.30%,and the recognition performance is better than that of a single CNN,with strong generalization ability and good robustness. |
| Key words: radar target recognition synthetic aperture radar(SAR) convolutional neural network(CNN) Ada_Nesterov momentum algorithm network ensemble |
