基于自适应动量估计优化器与空变最小熵准则的SAR图像船舶目标自聚焦算法

李志远; 郭嘉逸; 张月婷; 黄丽佳; 李洁; 吴一戎

doi:10.12000/JR21159

基于自适应动量估计优化器与空变最小熵准则的SAR图像船舶目标自聚焦算法

doi: 10.12000/JR21159

李志远^{1, 2, 3},
郭嘉逸^{1, 2, 3, ,},
张月婷^{1, 2, 3},
黄丽佳^{1, 2, 3},
李洁^{1, 2, 3},
吴一戎^{1, 2, 3}

1.
中国科学院空天信息创新研究院北京 100190
2.
中国科学院空间信息处理与应用系统技术重点实验室北京 100190
3.
中国科学院大学电子电气与通信工程学院北京 100049

基金项目: 国家自然科学基金(61991421)

详细信息

作者简介:
李志远(1997–)，男，山东人，中国科学院空天信息创新研究院博士研究生，2019年获北京理工大学信息与电子学院学士学位。主要研究方向为深度学习、SAR图像处理与图像质量提升等

郭嘉逸(1989–)，男，河北人，博士，现任中国科学院空天信息创新研究院助理研究员，研究方向为SAR图像深度学习应用、SAR图像与遥感图像复原应用，发表SCI论文10余篇

张月婷(1983–)，女，河北人，博士，现任中国科学院空天信息创新研究院副研究员，研究方向为SAR目标特性、SAR图像与遥感图像应用，发表SCI论文10余篇

黄丽佳(1984–)，女，辽宁人，博士，研究员。2011年在中国科学院电子学研究所获得博士学位，现任中国科学院空天信息创新研究院研究员。主要研究方向为合成孔径雷达信号处理与图像分析

李　洁(1996–)，女，山东人，中国科学院空天信息创新研究院博士研究生，2019年获北京理工大学信息与电子学院学士学位。主要研究方向为微波成像理论与技术、稀疏信号处理等

吴一戎(1963–)，男，安徽人，博士，中国科学院院士，博士生导师，国家杰出青年基金获得者，“百千万工程”国家级人选入选者。主要研究方向为微波成像理论与技术、雷达信号处理与雷达系统等

通讯作者:
郭嘉逸 jyguo@mail.ie.ac.cn

责任主编：杨淑媛 Corresponding Editor: YANG Shuyuan
中图分类号: TN958; TP391
计量
- 文章访问数: 1126
- HTML全文浏览量: 635
- PDF下载量: 130
- 被引次数: 0
出版历程
- 收稿日期: 2021-10-28
- 修回日期: 2022-01-05
- 网络出版日期: 2022-01-21
- 刊出日期: 2022-02-28

A Novel Autofocus Algorithm for Ship Targets in SAR Images Based on the Adaptive Momentum Estimation Optimizer and Space-variant Minimum Entropy Criteria

LI Zhiyuan^{1, 2, 3},
GUO Jiayi^{1, 2, 3
, ,},
ZHANG Yueting^{1, 2, 3},
HUANG Lijia^{1, 2, 3},
LI Jie^{1, 2, 3},
WU Yirong^{1, 2, 3}

1.
Aerospace Information Research Institute, Chinese Academy of Science, Beijing 100190, China
2.
Key Laboratory of Technology in Geo-spatial Information Processing and Application System, Chinese Acodemy of Sciences, Beijing 100190, China
3.
School of Electronic, Electrical and Communication Engineering, University of Chinese Academy of Sciences, Beijing 100049, China

Funds: The National Natural Science Foundation of China (61991421)

More Information

Corresponding author: GUO Jiayi, jyguo@mail.ie.ac.cn

摘要

摘要: 在SAR散焦船舶图像中，部分船舶目标的散焦现象具有沿距离向空变的特性。针对此类散焦船舶目标，该文提出了一种基于自适应动量估计优化器与空变最小熵准则的SAR图像船舶目标自聚焦算法，该算法直接对复图像进行处理，可以实现对任意阶次相位误差的补偿。在仿真数据和GF-3数据上的实验结果表明，所提算法可以有效地实现SAR图像空变散焦船舶目标自聚焦，聚焦后的船舶图像在图像熵与对比度上都有所改善，且算法聚焦速度有很大提升。
- 合成孔径雷达 /
- 船舶目标 /
- 自适应动量估计优化器 /
- 空变最小熵 /
- 自聚焦算法
Abstract: In SAR defocused ship images, the defocusing phenomenon of some ship targets is space-variant along the distance. In this context, a novel autofocus algorithm combining the adaptive momentum estimation optimizer and space-variant minimum entropy criteria is proposed to address these defocused ship targets. The algorithm can directly process complex images and compensate for any order phase errors. The effectiveness of the proposed method is proved by the experimental results on the simulation data and GF-3 data. Moreover, the entropy and contrast of the refocused image have been improved, and the focusing speed of the algorithm has been greatly enhanced.
- Synthetic Aperture Radar (SAR) /
- Ship target /
- Adaptive momentum estimation optimizer /
- Space-Variant Minimum Entropy (SVME) /
- Autofocus algorithm

HTML全文

图 1 添加不同阶次相位误差时方位回波信号压缩结果

Figure 1. The compression results of azimuth echo signal when adding different order phase error

下载: 全尺寸图片幻灯片

图 2 GF-3 空变散焦船舶目标

Figure 2. GF-3 space-variant defocused ship target

下载: 全尺寸图片幻灯片

图 3 算法流程图

Figure 3. Flowchart of the proposed algorithm

下载: 全尺寸图片幻灯片

图 4 空变散焦数据仿真算法流程图

Figure 4. Flowchart of space-variant defocused data simulation algorithm

下载: 全尺寸图片幻灯片

图 5 仿真空变散焦船舶图像

Figure 5. Simulated space-variant defocused ship images

下载: 全尺寸图片幻灯片

图 6 GF-3空变散焦船舶图像

Figure 6. GF-3 space-variant defocused ship images

下载: 全尺寸图片幻灯片

图 7 在仿真数据上的自聚焦结果(a1—d1：GF-3原始船舶图像，a2—d2：仿真的空变散焦船舶图像，a3—d3：算法1船舶图像聚焦结果，a4—d4：算法2船舶图像聚焦结果，a5—d5：算法3船舶图像聚焦结果，a6—d6：算法4船舶图像聚焦结果，a7—d7：算法5船舶图像聚焦结果)

Figure 7. Autofocus results on simulated data (a1—d1: the ground truth obtained in GF-3UFS mode, a2—d2: the simulated space-variant defocused ship images, a3—d3: the ship images refocused by algorithm 1, a4—d4: the ship images refocused by algorithm 2, a5—d5: the ship images refocused by algorithm 3, a6—d6: the ship images refocused by algorithm 4, a7—d7: the ship images refocused by algorithm 5)

下载: 全尺寸图片幻灯片

图 8 仿真数据自聚焦结果图像熵与对比度可视化

Figure 8. Visualization of image entropy and contrast of autofocus results on simulated data

下载: 全尺寸图片幻灯片

图 9 在GF-3数据上的自聚焦结果(a1—d1：GF-3超精细条带模式下获得的空变散焦船舶图像，a2—d2：算法1船舶图像聚焦结果，a3—d3：算法2船舶图像聚焦结果，a4—d4：算法3船舶图像聚焦结果，a5—d5：算法4船舶图像聚焦结果，a6—d6：算法5船舶图像聚焦结果)

Figure 9. Autofocus results on GF-3 data (a1—d1: the space-variant defocused ship images obtained in GF-3UFS mode, a2—d2: the ship images refocused by algorithm 1, a3—d3: the ship images refocused by algorithm 2, a4—d4: the ship images refocused by algorithm 3, a5—d5: the ship images refocused by algorithm 4, a6—d6: the ship images refocused by algorithm 5)

下载: 全尺寸图片幻灯片

图 10 GF-3数据自聚焦结果图像熵和对比度可视化

Figure 10. Visualization of image entropy and contrast of autofocus results on GF-3 data

下载: 全尺寸图片幻灯片

图 11 算法3中补偿的相位误差

Figure 11. The compensated phase error by algorithm 3

下载: 全尺寸图片幻灯片

表 1 使用Adam优化器估计相位误差系数伪代码

Table 1. Pseudo code for estimating phase error coefficients using Adam optimizer

输入：图像I 　输出：自聚焦图像$ f\left( {\boldsymbol I, \boldsymbol \alpha } \right) $
初始化：学习率$ \varepsilon $，矩估计的指数衰减速率$ {\gamma _{\text{1}}} $, $ {\gamma _{\text{2}}} $，相位误差系　　　　　数$\boldsymbol \alpha = \left[ { {\alpha _{\text{1} } },{\alpha _{\text{2} } },\cdots,{\alpha _{{i} } } } \right]$，用于数值稳定的小常数$ \delta $，梯度　　　　　1阶矩和2阶矩变量$ {\boldsymbol s}, {\boldsymbol r} $，迭代次数$ t $。
While 图像熵未收敛 do
计算图像熵梯度：$\boldsymbol g \leftarrow {\nabla _\alpha }L\left( {f\left( { {\boldsymbol I}, {\boldsymbol \alpha} } \right)} \right)$
更新有偏1阶矩估计：$ {\boldsymbol s} \leftarrow {\gamma _1}{\boldsymbol s} + \left( {1 - {\gamma _1}} \right){\boldsymbol g} $
更新有偏2阶矩估计：$ {\boldsymbol r} \leftarrow {\gamma _2}{\boldsymbol r} + \left( {1 - {\gamma _2}} \right){\boldsymbol g} \cdot {\boldsymbol g} $ (‘$ \cdot $’: 逐元素　　　相乘) 　　　修正1阶矩偏差：$\hat {\boldsymbol s}{\rm{ = } }\displaystyle\frac{\boldsymbol s}{ {1 - \gamma _1^t} }$
修正2阶矩偏差：$\hat {\boldsymbol r}{\rm{ = } }\displaystyle\frac{\boldsymbol r}{ {1 - \gamma _2^t} }$
更新相位误差系数：${\boldsymbol \alpha} = {\boldsymbol \alpha} - \displaystyle\frac{\varepsilon }{ {\sqrt {\hat {\boldsymbol r} } + \delta } }\hat {\boldsymbol s}$
end while

下载: 导出CSV

表 2 GF-3超精细条带图像参数

Table 2. The detailed information of GF-3 UFS SAR images

参数	值
成像模式	UFS
产品类型	SLC
产品级别	L1A
轨道模式	ASC
极化方式	HH/HV
地距分辨率(m)	2.5～5.0
方位向分辨率(m)	3
幅宽(km)	30
像元间距[Rg×Az](m)	1.124～1.728
入射角(°)	39.51～41.08

下载: 导出CSV

表 3 仿真数据上的算法速度对比

Table 3. Algorithm speed comparison on simulation data

算法	运行时间(s)
算法	图像a	图像b	图像c	图像d
算法1	101.35	81.25	96.25	94.35
算法2	32.16	21.53	32.45	32.49
算法3	2.31	1.73	1.75	2.36
算法4	0.69	0.69	0.57	0.68
算法5	0.25	0.23	0.25	0.18

下载: 导出CSV

表 4 GF-3数据上的算法速度对比

Table 4. Algorithm speed comparison on GF-3 data

算法	运行时间(s)
算法	图像a	图像b	图像c	图像d
算法1	92.25	81.11	205.56	170.38
算法2	10.73	10.73	42.39	32.28
算法3	1.75	1.76	4.61	2.88
算法4	0.71	0.69	0.70	0.73
算法5	0.23	0.24	0.23	0.24

下载: 导出CSV

参考文献(21)

[1]	NAKAMURA K, WAKABAYASHI H, NAOKI K, et al. Observation of sea-ice thickness in the sea of Okhotsk by using dual-frequency and fully polarimetric airborne SAR (pi-SAR) data[J]. IEEE Transactions on Geoscience and Remote Sensing, 2005, 43(11): 2460–2469. doi: 10.1109/TGRS.2005.853928
[2]	KRIEGER G, GEBERT N, and MOREIRA A. Unambiguous SAR signal reconstruction from nonuniform displaced phase center sampling[J]. IEEE Geoscience and Remote Sensing Letters, 2004, 1(4): 260–264. doi: 10.1109/LGRS.2004.832700
[3]	REIGBER A, SCHEIBER R, JAGER M, et al. Very-high-resolution airborne synthetic aperture radar imaging: Signal processing and applications[J]. Proceedings of the IEEE, 2013, 101(3): 759–783. doi: 10.1109/JPROC.2012.2220511
[4]	MOREIRA A, PRATS-IRAOLA P, YOUNIS M, et al. A tutorial on synthetic aperture radar[J]. IEEE Geoscience and Remote Sensing Magazine, 2013, 1(1): 6–43. doi: 10.1109/MGRS.2013.2248301
[5]	JING Wenbo, JIN Tian, and XIANG Deliang. Content-sensitive superpixel generation for SAR images with edge penalty and contraction-expansion search strategy[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 5210715. doi: 10.1109/TGRS.2021.3077407
[6]	NOVIELLO C, FORNARO G, and MARTORELLA M. Focused SAR image formation of moving targets based on Doppler parameter estimation[J]. IEEE Transactions on Geoscience and Remote Sensing, 2015, 53(6): 3460–3470. doi: 10.1109/TGRS.2014.2377293
[7]	SUN Huadong, ZHO Lei, MA Jinzhao, et al. The hybrid SAR-ISAR imaging algorithm applied to SAR moving target imaging[C]. The IEEE 9th International Conference on Fuzzy Systems and Knowledge Discovery, Chongqing, China, 2012: 1985–1988, doi: 10.1109/FSKD.2012.6234082.
[8]	邓云凯, 禹卫东, 张衡, 等. 未来星载SAR技术发展趋势[J]. 雷达学报, 2020, 9(1): 1–33. doi: 10.12000/JR20008 DENG Yunkai, YU Weidong, ZHANG Heng, et al. Forthcoming spaceborne SAR development[J]. Journal of Radars, 2020, 9(1): 1–33. doi: 10.12000/JR20008
[9]	CALLOWAY T M and DONOHOE G W. Subaperture autofocus for synthetic aperture radar[J]. IEEE Transactions on Aerospace and Electronic Systems, 1994, 30(2): 617–621. doi: 10.1109/7.272285
[10]	MANCILL C E and SWIGER J M. A map drift autofocus technique for correcting higher order SAR phase errors[C]. The 27th Annual Tri-Service Radar Symposium Record, Monterey, USA, 1981: 391–400.
[11]	侯育星, 邢孟道, 陈士超, 等. 稳健多子孔径图像偏移的SAR自聚焦算法[J]. 系统工程与电子技术, 2014, 36(9): 1731–1739. doi: 10.3969/j.issn.1001-506X.2014.09.10 HOU Yuxing, XING Mengdao, CHEN Shichao, et al. Robust autofocus algorithm of multiple subaperture mapdrift in SAR[J]. Systems Engineering and Electronics, 2014, 36(9): 1731–1739. doi: 10.3969/j.issn.1001-506X.2014.09.10
[12]	WAHL D E, EICHEL P H, GHIGLIA D C, et al. Phase gradient autofocus-a robust tool for high resolution SAR phase correction[J]. IEEE Transactions on Aerospace and Electronic Systems, 1994, 30(3): 827–835. doi: 10.1109/7.303752
[13]	BERIZZI F, CORSINI G, DIANI M, et al. Autofocus of wide azimuth angle SAR images by contrast optimisation[C]. 1996 International Geoscience and Remote Sensing Symposium, Lincoln, USA, 1996. doi: 10.1109/IGARSS.1996.516624.
[14]	ZENG Tao, WANG Rui, and LI Feng. SAR image autofocus utilizing minimum-entropy criterion[J]. IEEE Geoscience and Remote Sensing Letters, 2013, 10(6): 1552–1556. doi: 10.1109/LGRS.2013.2261975
[15]	武昕伟, 朱兆达. 一种基于最小熵准则的SAR图像自聚焦算法[J]. 系统工程与电子技术, 2003, 25(7): 867–869. doi: 10.3321/j.issn:1001-506X.2003.07.027 WU Xinwei and ZHU Zhaoda. A novel autofocus algorithm based on minimum entropy criteria for SAR images[J]. Systems Engineering and Electronics, 2003, 25(7): 867–869. doi: 10.3321/j.issn:1001-506X.2003.07.027
[16]	CARRARA W G, GOODMAN R S, and MAJEWSKI R M. Spotlight Synthetic Aperture Radar: Signal Processing Algorithms[M]. Boston: Artech House, 1995: 203–206
[17]	KINGMA D P and BA J. Adam: A method for stochastic optimization[C]. The 3rd International Conference on Learning Representations, San Diego, USA, 2015.
[18]	XI Li, LIU Guosui, and NI Jinlin. Autofocusing of ISAR images based on entropy minimization[J]. IEEE Transactions on Aerospace and Electronic Systems, 1999, 35(4): 1240–1252. doi: 10.1109/7.805442
[19]	KASILINGAM D, WANG Junfeng, LEE J S, et al. Focusing of synthetic aperture radar images of moving targets using minimum entropy adaptive filters[C]. IEEE 2000 International Geoscience and Remote Sensing Symposium on Taking the Pulse of the Planet: The Role of Remote Sensing in Managing the Environment, Honolulu, USA, 2000: 74–76. doi: 10.1109/IGARSS.2000.860426.
[20]	WU Haishan and BARBA J. Minimum entropy restoration of star field images[J]. IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics) , 1998, 28(2): 227–231. doi: 10.1109/3477.662762
[21]	马琳, 潘宗序, 黄钟泠, 等. 基于子孔径与全孔径特征学习的SAR多通道虚假目标鉴别[J]. 雷达学报, 2021, 10(1): 159–172. doi: 10.12000/JR20106 MA Lin, PAN Zongxu, HUANG Zhongling, et al. Multichannel false-target discrimination in SAR images based on sub-aperture and full-aperture feature learning[J]. Journal of Radars, 2021, 10(1): 159–172. doi: 10.12000/JR20106