[1]郑儒楠,刘文波.应用虚拟样本对SAR图像目标识别的研究[J].机械与电子,2017,(06):12-17.
 ZHENG Runan,LIU Wenbo.Research on SAR Image Target Recognition based on Virtual Sample[J].Machinery & Electronics,2017,(06):12-17.
点击复制

应用虚拟样本对SAR图像目标识别的研究
分享到:

《机械与电子》[ISSN:1001-2257/CN:52-1052/TH]

卷:
期数:
2017年06期
页码:
12-17
栏目:
设计与研究
出版日期:
2017-06-24

文章信息/Info

Title:
Research on SAR Image Target Recognition based on Virtual Sample
文章编号:
1001-2257(2017)06-0012-06
作者:
郑儒楠刘文波
(南京航空航天大学自动化学院,江苏 南京 211106)
Author(s):
ZHENG RunanLIU Wenbo
(College of Automation Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China)
关键词:
SAR图像 机器学习 目标识别 虚拟样本
Keywords:
SAR image machine learning object recognition virtual sample
分类号:
TP751.1
文献标志码:
A
摘要:
鉴于SAR图像获取的困难,无法保证机器学习算法时需要的大数据量训练样本,因此影响了识别结果。首次提出了应用虚拟样本来扩大SAR图像目标识别训练集,提高SAR 图像目标识别率的方法。通过使用重采样算法,奇异值重构与轮廓波重构等方法构建虚拟样本,与原有样本组成训练集,并通过SVM支持向量机进行训练识别在MSTAR公共数据集上的识别实验结果表明,对于不同数量的实际训练样本,通过添加本文方法构建的虚拟样本扩大训练集后,SAR 图像目标识别率均会得到提高,尤其在小样本的情况下,识别率提高非常显著。该文证实了虚拟样本应用于SAR图像目标识别的有效性。在样本数目有限时,添加虚拟样本对SAR图像目标识别性能具有明显的改善作用。
Abstract:
Machine learning algorithm is one of the main methods of recognizing ground target SAR image; however, because of the difficulties of SAR image acquisition, it is impossible to guarantee a sufficient number of training samples during the machine learning algorithm. Thus, the recognition results will be affected. For the first time, this paper proposed the way to improve the recognition rate of SAR images by using virtual samples to expand the training set of SAR image target recognition. Through such methods as re-sampling algorithm, singular value reconstruction and contourlet reconstruction to generate virtual samples, this paper combined the training set with the original samples. An experiment was conducted on recognizing MSTAR data sets by means of training vector machine with SVM support. The results show that for different numbers of training samples, the recognition rate of SAR images can be improved by adding the virtual samples, especially in the case of small samples. This paper proves the effectiveness of using virtual sample in SAR image target recognition. When the number of samples is limited, adding virtual samples can significantly improve the performance of SAR image target recognition.

参考文献/References:

[1] CAGATAY N D, Datcu M. Scene recognition based on phase gradient In SAR images[C]//Image Processing(ICIP), 2014 IEEE International Conference on. IEEE, 2014: 5162-5166.
[2] DING J, CHEN B, LIU H, et al. Convolutional Neural Network With Data Augmentation for SAR Target Recognition [J]. IEEE Geoscience and Remote Sensing Letters, 2016, 13(3): 364-368.
[3] DING S,NIE X, QIAO H, et al. Online classification for SAR target recognition based on SVM and approximate convex hull vertices selection[C]//Intelligent Control and Automation(WCICA), 2014 11th World Congress on. IEEE, 2014: 1473-1478.
[4] LI X, LI C, WANG P, et al. SAR ATR based on dividing CNN into CAE and SNN[C]//Synthetic Aperture Radar(APSAR), 2015 IEEE 5th Asia-Pacific Conference on. IEEE, 2015: 676-679.
[5] LI D C, WEN I H, CHANG CC. A new virtual-sample-generating method based on the heuristics algorithm[C]//Grey Systems and Intelligent Services, 2013 IEEE International Conference on. IEEE, 2013: 469-472.
[6] PORWAL S, Katiyar S K. Performance evaluation of various resampling techniques on IRS imagery[C]//Contemporary Computing(IC3), 2014 Seventh International Conference on. IEEE, 2014: 489-494.
[7] SUN X, MA T,JIN Y. SVD reconstruction algorithm in 3D SPECT imaging[C]//Nuclear Science Symposium Conference Record, 2004 IEEE. IEEE, 2004, 6: 3527-3530.
[8] TAO R, WAN H, WANG Y. Artifact-freedespeckling of SAR images using contourlet [J]. Geoscience and Remote Sensing Letters, IEEE, 2012, 9(5): 980-984.
[9] 朱亚辉, 彭国华. 基于奇异值分解的图像融合效果综合评价研究[J]. 西北工业大学学报, 2013, 31(1): 25-28.
[10] 刘嵩, 李时东, 郑明辉. 一种改进的基于奇异值扰动的单样本人脸识别方法[J]. 计算机工程与科学, 2012, 34(10): 88-91
[11] MIRI M S,Mahloojifar A. Retinal image analysis using curvelet transform and multistructure elements morphology by reconstruction[J]. Biomedical Engineering, IEEE Transactions on, 2011, 58(5): 1183-1192. [DOI: 10.1109/TBME.2010.2097599]
[12] MISHRA A K,MOTAUNG T. Application of linear and nonlinear PCA to SAR ATR[C]//Radioelektronika(RADIOELEKTRONIKA), 2015 25th International Conference. IEEE, 2015: 349-354.
[13] CHENG Y Q, LI H C,CELIK T, et al. FRFT-based improved algorithm of unsupervised change detection in SAR images via PCA and K-means clustering[C]//2013 IEEE International Geoscience and Remote Sensing Symposium-IGARSS. IEEE, 2013: 1952-1955.
[14] QU G, YU Q, WANG Y. An improved method for SAR image coastline detection based ondespeckling and SVM[C]//Radar Conference 2013, IET International. IET, 2013: 1-6.
[15] SU X, SANG H, YANG G. An SVM-based method for land and sea segmentation inpolarimetric SAR images[C]//Image and Signal Processing(CISP), 2011 4th International Congress on. IEEE, 2011, 3: 1205-1208.
[16] 宋婉莹, 李明, 吴艳, 等. 基于加权合成核与三重 Markov 场的极化 SAR 图像分类方法[J]. 电子学报, 2016,44(3): 520-526.
[17] 丁军, 刘宏伟, 王英华, 等. 一种联合阴影和目标区域图像的SAR目标识别方法 [J]. 电子与信息学报, 2015, 37(3):594-600.

备注/Memo

备注/Memo:
收稿日期:2017-02-08
基金项目:航空基金(20152052026); 国家自然科学基金(61471191)
作者简介:郑儒楠( 1992- ),男,江苏镇江人,硕士研究生,研究方向为计算机图形学、信号处理等; 刘文波( 1968- ),女,辽宁大连人,博士,教授,博士研究生导师,研究方向为信号处理及应用,计算机测试与控制技术。
更新日期/Last Update: 2017-06-25