基于多特征-多表示融合的SAR图像目标识别

张新征; 谭志颖; 王亦坚

doi:10.12000/JR17078

基于多特征-多表示融合的SAR图像目标识别

DOI: 10.12000/JR17078

重庆大学通信工程学院重庆 400044

基金项目: 国家自然科学基金(61301224)

详细信息

作者简介:
张新征(1978–)，男，山东省聊城人，博士，副教授。2009年于航天科工集团第二研究院获导航制导与控制专业博士学位，现担任重庆大学通信工程学院副教授。主要研究方向为遥感信息获取与处理、人工智能及其应用，目前已发表论文20余篇，专利获权5项。社会兼职：中国电子学会会员。E-mail: zhangxinzheng@cqu.edu.cn

谭志颖(1994–)，女，湖南郴州人，2016年在湘潭大学信息工程学院获得本科学位，现为重庆大学通信工程学院硕士研究生，研究方向为遥感图像分类。E-mail: 20161202031t@cqu.edu.cn

王亦坚(1994–)，男，江西抚州人，2016年在重庆大学通信工程学院获得本科学位，现为重庆大学通信工程学院硕士研究生，研究方向为SAR图像目标检测与识别。E-mail: 20161202029t@cqu.edu.cn

通讯作者:
张新征 zhangxinzheng@cqu.edu.cn

中图分类号: TN959
计量
- 文章访问数: 4323
- HTML全文浏览量: 890
- PDF下载量: 898
- 被引次数: 0
出版历程
- 收稿日期: 2017-08-18
- 修回日期: 2017-10-22
- 网络出版日期: 2017-10-28

SAR Target Recognition Based on Multi-feature Multiple Representation Classifier Fusion

College of Communication Engineering, Chongqing University, Chongqing 400044, China

Funds: The National Natural Science Foundation of China (61301224)

摘要

摘要: 针对合成孔径雷达(Synthetic Aperture Radar, SAR)图像目标识别问题，该文提出了一种基于多特征-多表示学习分类器融合的识别算法。首先，该算法提取了SAR图像3种特征，包括主成分(Principle Component Analysis, PCA)特征，小波变换特征和2维切片Zernike矩(2-Dimension Slice Zernike Moments, 2DSZM)特征。然后，将测试样本的3类特征分别输入稀疏表示分类器和协同表示分类器进行预分类，得到6个预测标签。对6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3种不同的分类器融合算法，实验结果表明利用贝叶斯决策融合得到了最佳的识别性能。基于多特征-多表示学习分类器融合的方法集成了多特征的鉴别能力，也融合了稀疏和协同表示的分类性能，实现优势互补，有效提高了识别精度。基于Moving and Stationary Target Acquisition and Recognition (MSTAR)公开发布的SAR目标数据库的实验验证了该方法的有效性。
- 合成孔径雷达 /
- 目标识别 /
- 稀疏表示 /
- 协同表示 /
- 决策融合
Abstract: In this paper, we present a Synthetic Aperture Radar (SAR) image target recognition algorithm based on multi-feature multiple representation learning classifier fusion. First, it extracts three features from the SAR images, namely principal component analysis, wavelet transform, and Two-Dimensional Slice Zernike Moments (2DSZM) features. Second, we harness the sparse representation classifier and the cooperative representation classifier with the above-mentioned features to get six predictive labels. Finally, we adopt classifier fusion to obtain the final recognition decision. We researched three different classifier fusion algorithms in our experiments, and the results demonstrate thatusing Bayesian decision fusion gives thebest recognition performance. The method based on multi-feature multiple representation learning classifier fusion integrates the discrimination of multi-features and combines the sparse and cooperative representation classification performance to gain complementary advantages and to improve recognition accuracy. The experiments are based on the Moving and Stationary Target Acquisition and Recognition (MSTAR) database,and they demonstrate the effectiveness of the proposed approach.
- Synthetic Aperture Radar (SAR) /
- Target recognition /
- Sparse representation /
- Collaborative representation /
- Decision fusion

HTML全文

图 1 3类目标的多层2维切片

Figure 1. Multiple 2D slices of three targets

下载: 全尺寸图片幻灯片

图 2 3类目标的2D-slice Zernike moments特征值比较

Figure 2. Comparisons of 2D-slice Zernike moments of three targets

下载: 全尺寸图片幻灯片

图 3 基于多特征-多表示学习分类器融合算法流程图

Figure 3. The procedure of based on fusion of multi-feature multiple representation classifier

下载: 全尺寸图片幻灯片

图 4 3类目标光学图像

Figure 4. Optical images of three targets

下载: 全尺寸图片幻灯片

图 5 3类目标的微波图像

Figure 5. Microwave images of three targets

下载: 全尺寸图片幻灯片

图 6 3种算法随特征维数变化的识别率

Figure 6. Recognition rate of three algorithms with feature dimension

下载: 全尺寸图片幻灯片

图 7 3种算法随规则化参数 $ λ $的识别率

Figure 7. Recognition rates of three algorithms with regularized parameters $ λ $

下载: 全尺寸图片幻灯片

图 8 SCBF随规则化参数 $ λ_1 $和 $ λ_2 $的识别率

Figure 8. Recognition rate of SCBF with the regularization parameters $ λ_1 $ and $ λ_2 $

下载: 全尺寸图片幻灯片

图 9 大俯仰角下3种算法的识别率柱状图

Figure 9. Histogram of recognition rate of three algorithms at large pitch angle

下载: 全尺寸图片幻灯片

表 1 实验数据集的型号/数目

Table 1. The types and numbers of training and testing datasets

目标	训练(17°)	测试(15°)
BMP2	sn-9563/233	sn-9563/195
	sn-9566/232	sn-9566/196
	sn-c21/233	sn-c21/196
BTR70	sn-c71/233	sn-c71/196
T72	sn-132/232	sn-132/196
	sn-812/231	sn-812/195
	sn-s7/228	sn-s7/191

下载: 导出CSV

表 2 3种算法随特征维数变化的识别率(%)

Table 2. Recognition rate of three algorithms with feature dimension (%)

特征维数	方法
特征维数	SCBF	SCMV	SCAWF
120	99.09	99.03	99.09
240	99.60	99.20	99.20
360	99.54	90.03	98.92
480	99.54	98.86	99.15
600	97.44	97.72	97.78

下载: 导出CSV

表 3 3种算法随规则化参数 $ λ $的识别率(%)

Table 3. Recognition rates of three algorithms with regularized parameters $ λ $ (%)

$ λ $	方法
$ λ $	MFSCF	MV	AWFA
10^–3	99.20	98.52	98.69
10^–2	99.15	98.52	98.75
10^–1	99.15	98.52	98.52
1	99.15	98.46	99.03
5	99.32	98.80	99.20
10	99.32	98.46	99.09

下载: 导出CSV

表 4 SCBF随规则化参数 $ λ $的混淆矩阵

Table 4. Confusion matrix of SCBF with regularized parameters $ λ $

$ λ $	识别结果	目标类型
$ λ $	识别结果	BMP2	BTR70	T72
10^–3	BMP2	580	4	3
	BTR70	4	584	0
	T72	3	0	579
10^–2	BMP2	580	4	4
	BTR70	4	584	0
	T72	3	0	578
10^–1	BMP2	581	5	3
	BTR70	3	583	1
	T72	3	0	578
1	BMP2	579	4	3
	BTR70	5	584	0
	T72	3	0	579
5	BMP2	579	3	1
	BTR70	4	585	0
	T72	4	0	581
10	BMP2	579	3	1
	BTR70	4	585	0
	T72	4	0	581

下载: 导出CSV

表 5 大俯仰角实验中使用的数据集

Table 5. Dataset used in large depression angle experiment

目标	Training set (17°)	Testing set (30°)	Testing set (45°)
BRDM2	298	287	303
2S1	299	288	303
ZSU234	299	288	303

下载: 导出CSV

表 6 大俯仰角下不同分类方法的识别率(%)

Table 6. Recognition rate for different classification methods at large pitch angles (%)

俯仰角 (°)	分类方法
俯仰角 (°)	SRC (PCA)	SRC (小波)	SRC (2DSZM)	CRC (PCA)	CRC (小波)	CRC (2DSZM)	SCMV	AFWF	SCBF
30	86.91	33.49	87.60	84.13	32.79	88.30	86.56	90.38	94.79
45	63.37	36.63	44.44	59.96	33.33	45.32	43.78	46.86	68.65

下载: 导出CSV

参考文献(25)

[1]	Qi Zhi-xin, Yeh A G O, Li Xia, et al.. A novel algorithm for land use and land cover classification using RADARSAT-2 polarimetric SAR data[J]. Remote Sensing of Environment, 2012, 118: 21–39. doi: 10.1016/j.rse.2011.11.001
[2]	Liu Bin, Hu Hao, Wang Huan-yu, et al.. Superpixel-based classiﬁcation with an adaptive number of classes for polarimetric SAR images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2013, 51(2): 907–924. doi: 10.1109/TGRS.2012.2203358
[3]	Wang Hui, Chen Zhan-sheng, and Zheng Shi-chao. Preliminary research of Low-RCS moving target detection based on Ka-Band video SAR[J]. IEEE Geoscience and Remote Sensing Letters, 2017, 14(6): 811–815. doi: 10.1109/LGRS.2017.2679755
[4]	El-DarymliK, Gill E W, and Mcguire P. Automatic target recognition in synthetic aperture radar imagery: A state-of-the-art review[J]. IEEE Access, 2016, 4: 6014–6058. doi: 10.1109/ACCESS.2016.2611492
[5]	Novak L M, Owirka G J, and Netishen C M. Performance of a high-resolution polarimetric SAR automatic target recognition system[J]. The Lincoln Laboratory Journal, 1993, 6(1): 11–23.
[6]	Saghri J A and DeKelaita A. Exploitation of target shadows in synthetic aperture radar imagery for automatic target recognition[C]. Proceedings of SPIE Volume 6312 Applications of Digital Image Processing XXIX, California, United States, 2006, 6312: 631212. DOI: 10.1117/12.684401.
[7]	Amoon M and Rezai-Rad G A. Automatic target recognition of synthetic aperture radar (SAR) images based on optimal selection of Zernike moments features[J]. IET Computer Vision, 2014, 8(2): 77–85. doi: 10.1049/iet-cvi.2013.0027
[8]	Gerry M J, Potter L C, Gupta I J, et al.. A parametric model for synthetic aperture radar measurements[J]. IEEE Transactions on Antennas and Propagation, 1999, 47(7): 1179–1188. doi: 10.1109/8.785750
[9]	宦若虹, 张平, 潘赟. PCA、ICA和Gabor小波决策融合的SAR目标识别[J]. 遥感学报, 2012, 16(2): 262–274. doi: 10.11834/jrs.20120457 Huan Ruo-hong, Zhang Ping, and Pan Yun. SAR target recognition using PCA, ICA and Gabor wavelet decision fusion[J]. Journal of Remote Sensing, 2012, 16(2): 262–274. doi: 10.11834/jrs.20120457
[10]	Lin Chang, Peng Fei, Wang Bing-hui, et al.. Research on PCA and KPCA self-fusion based MSTAR SAR automatic target recognition algorithm[J]. Journal of Electronic Science and Technology, 2012, 10(4): 352–357.
[11]	Zhang Zheng, Xu Yong, Yang Jian, et al.. A survey of sparse representation: Algorithms and applications[J]. IEEE Access, 2017, 3: 490–530. doi: 10.1109/ACCESS.2015.2430359
[12]	Zhang Hai-chao, Nasrabadi N, Zhang Yan-ning, et al.. Multi-view automatic target recognition using joint sparse representation[J]. IEEE Transactions on Aerospace and Electronic Systems, 2012, 48(3): 2481–2497. doi: 10.1109/TAES.2012.6237604
[13]	Dong Gang-gang, Kuang Gang-yao, Wang Na, et al.. SAR target recognition via Joint sparse representation of monogenicsignal[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2015, 8(7): 3316–3328. doi: 10.1109/JSTARS.2015.2436694
[14]	Dong Gang-gang and Kuang Gang-yao. SAR target recognition via sparse representation of Monogenic signal on Grassmann manifolds[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2016, 9(3): 1308–1319. doi: 10.1109/JSTARS.2015.2513481
[15]	Sun Yong-gang, Du Lan, Wang Yan, et al.. SAR automatic target recognition based on dictionary learning and joint dynamic sparse representation[J]. IEEE Geoscience and Remote Sensing Letters, 2016, 13(12): 1777–1781. doi: 10.1109/LGRS.2016.2608578
[16]	Song Sheng-li, Xu Bin, and Yang Jian. SAR target recognition via supervised discriminative dictionary learning and sparse representation of the SAR-HOG feature[J]. Remote Sensing, 2016, 8(8): 683. doi: 10.3390/rs8080683
[17]	Liu Hong-wei, Bo Jiu, Li Fei, et al.. Attributed scattering center extraction algorithm based on sparse representation with dictionary refinement[J]. IEEE Transactions on Antennas and Propagation, 2017, 65(5): 2604–2614. doi: 10.1109/TAP.2017.2673764
[18]	Zhang Lei, Yang Meng, and Feng Xiang-chu. Sparse representation or collaborative representation: Which helps face recognition?[C]. Proceedings of IEEE International Conference on Computer Vision, Barcelona, Spain, 2012: 471–478.
[19]	Li Wei, Du Qian, Zhang Fan, et al.. Hyperspectral image classification by fusing collaborative and sparse representations[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2016, 9(9): 4178–4187. doi: 10.1109/JSTARS.2016.2542113
[20]	Chi Yue-jie and Porikli F. Classification and Boosting with multiple collaborative representations[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014, 36(8): 1519–1531. doi: 10.1109/TPAMI.2013.236
[21]	Haghighi M S, Vahedian A, and Yazdi H S. Extended decision template presentation for combining classifiers[J]. Expert Systems with Applications, 2011, 38(7): 8414–8418. doi: 10.1016/j.eswa.2011.01.036
[22]	Liu Ming, Wu Yan, Zhao Wei, et al.. Dempster-Shafer fusion of multiple sparse representation and statistical property for SAR target configuration recognition[J]. IEEE Geoscience and Remote Sensing Letters, 2014, 11(6): 1106–1109. doi: 10.1109/LGRS.2013.2287295
[23]	Liu Hai-cang and Li Shu-tao. Decision fusion of sparse representation and support vector machine for SAR image target recognition[J]. Neurocomputing, 2013, 113: 97–104. doi: 10.1016/j.neucom.2013.01.033
[24]	Zhang Xin-zheng, Liu Zhou-ying, Liu Shu-jun, et al.. Sparse coding of 2D-slice Zernike moments for SAR ATR[J]. International Journal of Remote Sensing, 2017, 38(2): 412–431. doi: 10.1080/01431161.2016.1266107
[25]	Xu Yong and Lu Yuwu. Adaptive weighted fusion: A novel fusion approach for image classification[J]. Neurocomputing, 2015, 168: 566–574. doi: 10.1016/j.neucom.2015.05.070