系统工程与电子技术 ›› 2018, Vol. 40 ›› Issue (11): 2433-.doi: 10.3969/j.issn.1001-506X.2018.11.07

• 传感器与信号处理 • 上一篇    下一篇

基于卷积稀疏编码与多分类器融合的雷达HRRP目标识别方法

王彩云1, 胡允侃1, 李晓飞2, 魏文怡1, 赵焕玥1   

  1. 1. 南京航空航天大学航天学院, 江苏 南京 210016;
    2. 北京电子工程总体研究所, 北京 100854
  • 出版日期:2018-10-25 发布日期:2018-11-14

Radar HRRP target recognition based on convolutional sparse coding and multi-classifier fusion

WANG Caiyun1, HU Yunkan1, LI Xiaofei2, WEI Wenyi1, ZHAO Huanyue1   

  1. 1. College of Astronautics, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China; 2. Beijing Institute of Electronic System Engineering, Beijing 100854, China
  • Online:2018-10-25 Published:2018-11-14

摘要:

针对雷达高分辨距离像(high resolution range profile, HRRP)目标识别问题,提出基于卷积稀疏编码与多分类融合(convolutional sparse coding and multi-classifier fusion,CSCMF)的识别方法。首先,该方法利用CSC方法对目标HRRP进行特征提取,同时实现数据压缩;然后,将测试样本的特征分别输入随机森林分类器、朴素贝叶斯分类器和最小值分类器进行预分类,得到3个预测标签。采用多数投票法对3个预测标签进行分类器融合,得到最终的识别决策。实验中研究了分类器融合方法。基于5种飞机目标的HRRP仿真数据进行了实验验证,实验结果表明该方法的分类准确率较高,而且对噪声有较强的鲁棒性。

Abstract:

A radar high resolution range profile (HRRP) target recognition algorithm based on convolutional sparse coding and classifier fusion method, named convolutional sparse coding and multi-classifier fusion (CSCMF) is proposed. Firstly, it extracts the features from the HRRPs using the convolutional sparse coding (CSC) method, and realizes the compression of the data set. Secondly, three different classifiers (random forest classifier, naive Bayesian classifier, and minimum classifier) that fuse the sparse coding characteristics were used to obtain three predictive labels. Finally, we adopt classifier fusion by the majority of the voting methods to get the final recognition decision. We researched some classifiers algorithms in our experiments, and the simulation results based on the radar high resolution range profile database demonstrate the presented method can achieve remarkable classification performance and more robust to noise.