计算机与现代化 ›› 2023, Vol. 0 ›› Issue (01): 37-42.

• 人工智能 • 上一篇    下一篇

基于多卷积神经网络融合的SAR舰船分类

  

  1. (1. 中国科学院空天信息创新研究院,北京  100190; 2. 中国科学院大学电子电气与通信工程学院,北京  100049)
  • 出版日期:2023-03-02 发布日期:2023-03-02
  • 作者简介:张骁(1996—),男,河北邯郸人,硕士研究生,研究方向:SAR图像舰船检测与分类,E-mail: zhangxiao19@mails.ucas.ac.cn; 通信作者:吕继宇(1976—),女,河南驻马店人,博士研究生,研究方向:星载合成孔径雷达系统设计与仿真,E-mail:jylv@mail.ie.ac.cn; 赵爽(1995—),女,吉林长春人,博士研究生,研究方向:SAR图像目标检测,E-mail: zhaoshuang19@mails.ucas.ac.cn; 吴羽纶(1997—),男,河北廊坊人,博士研究生,研究方向:SAR图像处理与InSAR处理,E-mail: wuyulun19@mails.ucas.ac.cn; 王春乐(1986—),女,吉林吉林人,博士研究生,研究方向:合成孔径雷达图像处理,E-mail: clwang@mail.ie.ac.cn。
  • 基金资助:
    国家自然科学基金资助项目(61901445)

SAR Ship Classification Based on Multi-convolutional Neural Network Fusion

  1. (1. Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100190, China; 2. School of Electronic, Electrical and Communication Engineering, University of Chinese Academy of Sciences, Beijing 100049, China)
  • Online:2023-03-02 Published:2023-03-02

摘要: 针对SAR图像中小型舰船分类准确率较低的问题,提出一种多卷积神经网络加权融合的方法。首先构建高分辨率卷积神经网络对特征图进行多尺度融合,引入微调模型和标签平滑减少训练过拟合的问题;然后利用高分辨网络、MobileNetv2网络和SqueezeNet网络训练3种单分类模型;最后采用加权投票方式对3种分类模型的结果进行融合。采用融合算法对GF-3号舰船数据集进行分类实验,取得94.83%的准确率、95.43%的召回率和0.9513的F1分数的分类性能。实验结果表明,该舰船分类算法模型具有较优的分类能力,验证了其在高分辨率SAR图像舰船分类上的有效性。

关键词: SAR图像, 高分辨率卷积神经网络, 微调模型, 标签平滑, 加权投票, 舰船分类

Abstract: The accuracy of small ship classification in Syntactic Aperture Radar (SAR) images is low. To solve the problem, a classification approach based on the weighted fusion of different convolutional neural network results is proposed. Firstly, a high-resolution convolutional neural network is constructed to conduct multi-scale feature fusion, fine-tuning model and label smoothing are introduced to reduce the problem of training over-fitting. Then three single classification models are trained using the high-resolution network, MobileNetv2 network and SqueezeNet network. Finally, the results of three classification models are fused by weighted voting. The fusion method is used to carry out classification experiment on GF-3 ship dataset, the results obtained are: precision 94.83%, recall rate 95.43%, F1 score 0.9513. Experimental results show that the algorithm model proposed in this paper has better classification ability, which verifies its effectiveness in high-resolution SAR image ship classification.

Key words: SAR images; high-resolution convolutional neural networks; fine tuning model, label smooth; weighted voting; ship classification