| 摘要: |
| 传统雷达高分辨一维距离像(High-resolution Range Profile,HRRP)目标识别方法只利用目标幅度信息而丢失其相位信息，这势必会造成信息不完备。为解决此问题，提出将深度极限学习机从实数域扩展到复数域，以有效提取复HRRP序列的深层潜在结构信息。同时为更好地保持数据间的邻域信息，将流形正则化引入到网络模型训练过程中，提出流形正则深度复极限学习机。在雷达暗室测量数据上的实验结果表明，所提算法相比常用的深度学习模型具有更好的识别效果和更快的训练速度，验证了算法的有效性。 |
| 关键词: 雷达目标识别 HRRP目标 极限学习机 深度学习 流形正则化 |
| DOI： |
|
| 基金项目: |
|
| Application of deep complex extreme learning machine in radar HRRP target recognition |
| ZHAO Feixiang,DU Jun,LIU Heng,MA Zilong |
| (China Huayin Ordnance Test Center,Huayin 714200,China) |
| Abstract: |
| The traditional radar high-resolution range profile(HRRP) target recognition method only uses the target amplitude information and loses its phase information,which will inevitably cause incomplete information.In order to solve this problem,this paper proposes to extend the deep extreme learning machine from the real domain to the complex domain to effectively extract the deep potential structural information of the complex HRRP sequence.At the same time,in order to better maintain the neighborhood relationships between data,manifold regularization is introduced into the training process of network model,and the manifold regularization complex deep extreme learning machine is proposed.Experimental results on radar darkroom measurement data show that the proposed algorithm has better recognition effect and faster training speed than commonly-used deep learning models,which verifies the effectiveness of the algorithm. |
| Key words: radar target recognition HRRP target extreme learning machine deep learning manifold regularization |
