系统工程与电子技术 ›› 2021, Vol. 43 ›› Issue (12): 3533-3541.doi: 10.12305/j.issn.1001-506X.2021.12.15

• 传感器与信号处理 • 上一篇    下一篇

基于一维堆叠池化融合卷积自编码器的HRRP目标识别方法

张国令1, 吴崇明2,*, 李睿1, 来杰1, 向前1   

  1. 1. 空军工程大学防空反导学院, 陕西 西安 710051
    2. 西京学院, 陕西 西安 710123
  • 收稿日期:2020-06-28 出版日期:2021-11-24 发布日期:2021-11-30
  • 通讯作者: 吴崇明
  • 作者简介:张国令(1995—), 男, 硕士研究生, 主要研究方向为人工智能、智能信息处理|吴崇明(1966—), 男, 副教授, 博士, 主要研究方向人工智能、智能信息处理|李睿(1992—), 男, 博士研究生, 主要研究方向为人工智能、智能信息处理|来杰(1994—), 男, 博士研究生, 主要研究方向为人工智能、网络空间安全|向前(1995—), 男, 博士研究生, 主要研究方向为人工智能、智能信息处理、网络空间安全
  • 基金资助:
    国家自然科学基金(61876189);国家自然科学基金(61273275);国家自然科学基金(61503407);国家自然科学基金(61806219);国家自然科学基金(61703426)

HRRP target recognition method based on one-dimensional stacked pooling fusion convolutional autoencoder

Guoling ZHANG1, Chongming WU2,*, Rui LI1, Jie LAI1, Qian XIANG1   

  1. 1. Air and Missile Defense College, Air Force Engineering University, Xi'an 710051, China
    2. Xijing College, Xi'an 710123, China
  • Received:2020-06-28 Online:2021-11-24 Published:2021-11-30
  • Contact: Chongming WU

摘要:

针对雷达高分辨距离像(high resolution range profile, HRRP)目标识别中特征提取及识别问题, 提出了一种基于一维堆叠池化融合卷积自编码器(one-dimensional stacked pooling fusion convolutional autoencoder, 1D SPF-CAE)的识别方法。首先构造一维池化融合卷积自编码器(one-dimensional pooling fusion convolutional auto-encoder, 1D PF-CAE), 在编码阶段, 采用最大池化和平均池化同时提取不同的编码特征并进行融合来提取HRRP的结构特征; 然后堆叠多个1D PF-CAE形成1D SPF-CAE; 最后使用标签数据对网络进行微调, 实现HRRP目标识别。并使用AdaBound算法优化网络训练来提高识别性能。基于弹道中段目标仿真数据的实验结果表明, 该方法具有较强的特征提取能力, 对于HRRP目标识别准确率高、鲁棒性强。

关键词: 雷达自动目标识别, 高分辨距离像, 卷积自编码器, 特征提取, 池化融合

Abstract:

Aiming at the problem of feature extraction and recognition in high resolution range profile (HRRP) target recognition, a recognition method based on one-dimensional stacked pooling fusion convolutional autoencoder (1D SPF-CAE) is proposed in this paper. Firstly, a one-dimensional pooling fusion convolutional autoencoder (1D PF-CAE) is constructed. In the encoding stage, the maximum pooling and average pooling are used to extract different encoding features and fuse them to extract the structural features of HRRP. Then, multiple 1D PF-CAEs are stacked to form 1D SPF-CAE. Finally, the network is fine-tuned using label data to realize HRRP target recognition. And the AdaBound algorithm is used to optimize network training for improving the recognition performance. The experimental results based on the simulated data of the target in the middle part of the trajectory show that the method has strong feature extraction capability, and has high accuracy and robustness for HRRP target recognition.

Key words: radar automatic target recognition, high resolution range profile, convolution autoencoder, feature extraction, pooling fusion

中图分类号: