基于局部超分辨重建的高精度SAR图像水域分割方法

李宁; 牛世林

doi:10.12000/JR19096

基于局部超分辨重建的高精度SAR图像水域分割方法

DOI: 10.12000/JR19096

李宁^1,2,3, ,,
牛世林^1,

1.
河南大学计算机与信息工程学院开封 475004
2.
河南省智能技术与应用工程技术研究中心开封 475004
3.
河南省大数据分析与处理重点实验室开封 475004

基金项目: 国家自然科学基金(U1604145, 61871175, 61601437)，河南省高等学校重点科研项目(18B520010, 19A420005)，河南省科技攻关计划项目(182102210233, 192102210082)，河南省青年人才托举工程(2019HYTP006)，河南大学研究生教育创新与质量提升计划项目(SYL18060127)

详细信息

作者简介:
李　宁(1987–)，男，安徽人，毕业于中国科学院电子学研究所，获得博士学位，现为河南大学教授，研究方向为多模式合成孔径雷达成像及其应用技术。E-mail: lining_nuaa@163.com

牛世林(1993–)，男，河南人，河南大学计算机与信息工程学院硕士研究生，主要研究方向为合成孔径雷达图像处理及其应用技术。E-mail: nsl1993@foxmail.com

通讯作者:
李宁 lining_nuaa@163.com

中图分类号: TN959.1; TP183
计量
- 文章访问数: 4146
- HTML全文浏览量: 1153
- PDF下载量: 370
- 被引次数: 19
出版历程
- 收稿日期: 2019-11-06
- 修回日期: 2020-02-02
- 网络出版日期: 2020-02-28

High-precision Water Segmentation from Synthetic Aperture Radar Images Based on Local Super-resolution Restoration Technology

LI Ning^{1,2,3
, ,},
NIU Shilin^1
,

1.
College of Computer and Information Engineering, Henan University, Kaifeng 475004, China
2.
Henan Engineering Research Center of Intelligent Technology and Application, Kaifeng 475004, China
3.
Henan Key Laboratory of Big Data Analysis and Processing, Henan University, Kaifeng 475004, China

Funds: The National Natural Science Foundation of China (U1604145, 61871175, 61601437), The College Key Research Project of Henan Province (18B520010, 19A420005), The Plan of Science and Technology of Henan Province (182102210233, 192102210082), The Youth Talent Lifting Project of Henan Province (2019HYTP006), The Graduate Education Innovation and Quality Improvement Program of henan University (SYL18060127)

More Information

Corresponding author: LI Ning, lining_nuaa@163.com

摘要

摘要: 合成孔径雷达(SAR)图像水域分割在水资源调查、灾害监测等领域具有重要意义。针对中低分辨率星载SAR图像水域提取精度不足的难题，该文融合基于轻量级残差卷积神经网络(CNN)的图像超分辨率重建技术和传统SAR图像水域分割技术的优点，提出了一种基于局部超分辨重建的SAR图像水域分割方法，显著提升了SAR图像水域分割的精度。为了验证上述方法的有效性，该文以南水北调中线工程水源地丹江口水库为应用对象，基于国产高分三号(GF-3)卫星的8 m分辨率标准条带(SS)模式图像和欧空局Sentinel-1卫星20 m分辨率干涉宽幅(IW)模式图像，开展了水域分割的实验验证和精度评估工作。实验结果表明，该文所提方法可在中低分辨率SAR图像中获取更精确的水域分割结果，其水域分割性能较传统方法有大幅提升。
- 合成孔径雷达 /
- 水域分割 /
- 卷积神经网络 /
- 超分辨率重建
Abstract: The extraction of water from Synthetic Aperture Radar (SAR) images is of great significance in water resources investigation and monitoring disasters. To deal with the problems of the insufficient accuracy of water boundaries extracted from middle-low resolution SAR images. This paper proposes a high-precision water boundaries extraction method based on a local super-resolution restoration technology that combines the advantages of the super-resolution restoration technology based on the lightweight residual Convolutional Neural Network (CNN) and the traditional SAR images water extraction methods. The proposed method can significantly improve the accuracy of water segmentation results by using SAR images. To verify the effectiveness of the proposed method, as a study area, we selected the Danjiangkou Reservoir, the water source of the middle route of a south-to-north water diversion project. Further, we conducted experiments on the multi-mode SAR dataset and evaluated its accuracy. This dataset included one Standard Strip-map (SS) mode image obtained by the Chinese GaoFen-3 (GF-3) satellite with a resolution of 8 m and one Interferometric Wide-swath (IW) mode SAR image obtained by Sentinel-1 satellite with a resolution of 20 m. The experimental results showed that the water segmentation results from the middle–low resolution SAR images of the proposed method were more precise, and the overall water segmentation performance was superior to that of the traditional methods.
- Synthetic Aperture Radar (SAR) /
- Water segmentation /
- Convolutional Neural Network (CNN) /
- Super-resolution restoration

HTML全文

图 1 基于局部超分辨重建的SAR图像高精度水域分割方法示意图

Figure 1. The structure of local SR based high-precision water extraction method for SAR image

下载: 全尺寸图片幻灯片

图 2 局部分块方法示意图

Figure 2. Schematic diagram of local block method

下载: 全尺寸图片幻灯片

图 3 LRSR网络结构图

Figure 3. The structure of the LRSR network

下载: 全尺寸图片幻灯片

图 4 精细分割水域边界融合方法

Figure 4. Merging method of the refined water boundaries

下载: 全尺寸图片幻灯片

图 5 SAR图像3倍超分辨重建结果

Figure 5. SR result of SAR images with upscaling factor of 3

下载: 全尺寸图片幻灯片

图 6 研究区域

Figure 6. Location of study area

下载: 全尺寸图片幻灯片

图 7 SAR图像水域分割结果

Figure 7. Water extraction results of SAR images

下载: 全尺寸图片幻灯片

表 1 LRSR网络卷积与反卷积参数设置

Table 1. Convolution and deconvolution parameters of the LRSR

数据处理层[卷积数目：深度× ( ${I_i}$ × ${O_i}$ )，卷积核大小：( ${F_i}$ × ${F_i}$ )]	padding	stride
特征提取层 [(1×64), (5×5)]	0	1
特征压缩层 [(64×16), (1×1)]	0	1
特征映射层 [m×(16×16), (3×3)]	1	1
特征扩张层 [(1×64), (1×1)]	0	1
超分辨重建层 [(64×1), (9×9)]	4	k

下载: 导出CSV

表 2 LRSR网络训练参数

Table 2. Convolution and deconvolution parameters of the LRSR

卷积学习率	卷积偏置学习率	反卷积学习率	反卷积偏置学习率	权值衰减	最大迭代次数
10^–3	10^–4	10^–4	2 × 10^–4	10^–4	10^–7

下载: 导出CSV

表 3 LRSR网络MSE损失训练结果

Table 3. LRSR training results of MSE loss

k	m
k	4	5	6	7	8
2	0.01767	0.01725	0.01754	0.01847	0.01894
3	0.09077	0.08939	0.08856	0.08866	0.08978
4	0.25842	0.25220	0.25034	0.25639	0.25464

下载: 导出CSV

表 4 实验使用的SAR图像详细参数

Table 4. Detailed parameters of experimental SAR data

	SAR图像1	SAR图像2
卫星	Sentinel-1	GF-3
成像模式	干涉宽测绘带(Interferometric Wide-swath, IW)	标准条带 (Stand Stripmap, SS)
成像日期	2019年2月8日	2017年7月14日
标称分辨率	20 m	8 m
图像尺寸	3824×2255像素	7428×5221像素
极化方式	VV	HH

下载: 导出CSV

表 5 对比试验方法

Table 5. Comparison test methods

对比方法	方法详情
方法1	水域粗分割(FCM聚类)+ACM精细分割
方法2	水域粗分割(相干斑滤波+FCM聚类+ROI提取)+ACM精细分割
方法3	水域粗分割(相干斑滤波+FCM聚类+ROI提取)+水域边界局部分块法+ACM精细分割
本文方法	水域粗分割(相干斑滤波+FCM聚类+ROI提取)+水域边界局部分块法+图像超分辨重建+ACM精细分割

下载: 导出CSV

表 6 定量分析结果

Table 6. Results of quantitative analysis

SAR图像编号	对比方法1			对比方法2			对比方法3			本文方法
SAR图像编号	虚警率(%)	准确率(%)	轮廓平均偏移像素(个)	虚警率(%)	准确率(%)	轮廓平均偏移像素(个)	虚警率(%)	准确率(%)	轮廓平均偏移像素(个)	虚警率(%)	准确率(%)	轮廓平均偏移像素(个)
1	20.43	85.23	2.1714	10.59	91.06	1.5411	0.68	98.69	0.7957	0.014	99.65	0.1543
2	9.27	90.15	1.6228	3.28	92.43	1.3652	0.91	98.83	0.7468	0.034	99.72	0.1403

下载: 导出CSV

参考文献(19)

[1]	牛世林, 郭拯危, 李宁, 等. 星载SAR水域分割研究进展与趋势分析[J]. 聊城大学学报: 自然科学版, 2018, 118(2): 75–89. NIU Shilin, GUO Zhengwei, LI Ning, et al. Research progress and trend analysis of water extraction by spaceborne SAR[J]. Journal of Liaocheng University:Natural Science Edition, 2018, 118(2): 75–89.
[2]	LI Ning, NIU Shilin, GUO Zhengwei, et al. Dynamic waterline mapping of inland great lakes using time-series SAR data from GF-3 and S-1A satellites: A case study of DJK reservoir, China[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2019, 12(11): 4297–4314. doi: 10.1109/JSTARS.2019.2952902
[3]	邓云凯, 赵凤军, 王宇. 星载SAR技术的发展趋势及应用浅析[J]. 雷达学报, 2012, 1(1): 1–10. doi: 10.3724/SP.J.1300.2012.20015 DENG Yunkai, ZHAO Fengjun, and WANG Yu. Brief analysis on the development and application of spaceborne SAR[J]. Journal of Radars, 2012, 1(1): 1–10. doi: 10.3724/SP.J.1300.2012.20015
[4]	GUO Yaru and ZHANG Jixian. A new 2D OTSU for water extraction from SAR image[J]. ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2017, XLII-2/W7: 733–736. doi: 10.5194/isprs-archives-XLII-2-W7-733-2017
[5]	LI Ning, WANG R, DENG Yunkai, et al. Waterline mapping and change detection of Tangjiashan dammed lake after Wenchuan earthquake from multitemporal high-resolution airborne SAR imagery[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2014, 7(8): 3200–3209. doi: 10.1109/JSTARS.2014.2345417
[6]	冷英, 刘忠玲, 张衡, 等. 一种改进的ACM算法及其在鄱阳湖水域监测中的应用[J]. 电子与信息学报, 2017, 39(5): 1064–1070. doi: 10.11999/JEIT160870 LENG Ying, LIU Zhongling, ZHANG Heng, et al. Improved ACM algorithm for Poyang lake monitoring[J]. Journal of Electronics &Information Technology, 2017, 39(5): 1064–1070. doi: 10.11999/JEIT160870
[7]	冷英, 李宁. 一种改进的变化检测方法及其在洪水监测中的应用[J]. 雷达学报, 2017, 6(2): 204–212. doi: 10.12000/JR16139 LENG Ying and LI Ning. Improved change detection method for flood monitoring[J]. Journal of Radars, 2017, 6(2): 204–212. doi: 10.12000/JR16139
[8]	张金松, 邢孟道, 孙光才. 一种基于密集深度分离卷积的SAR图像水域分割算法[J]. 雷达学报, 2019, 8(3): 400–412. doi: 10.12000/JR19008 ZHANG Jinsong, XING Mengdao, and SUN Guangcai. A water segmentation algorithm for SAR image based on dense depthwise separable convolution[J]. Journal of Radars, 2019, 8(3): 400–412. doi: 10.12000/JR19008
[9]	SZILAGYI L, BENYÓ Z, SZILAGYI S M, et al. MR brain image segmentation using an enhanced fuzzy C-means algorithm[C]. The 25th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Cancun, Mexico, 2015, 1: 724–726. doi: 10.1109/IEMBS.2003.1279866.
[10]	CHEN Jiaqi, WANG Qingwei, WANG Jian, et al. Change detection of water index in danjiangkou reservoir using mixed log-normal distribution based active contour model[J]. IEEE Access, 2019, 7: 95430–95442. doi: 10.1109/ACCESS.2019.2929178
[11]	DONG Chao, LOY C C, HE Kaiming, et al. Image super-resolution using deep convolutional networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014, 38(2): 295–307. doi: 10.1109/TPAMI.2015.2439281
[12]	DONG Chao, LOY C C, and TANG Xiaoou. Accelerating the super-resolution convolutional neural network[C]. The 14th European Conference on Computer Vision (ECCV’16), Amsterdam, The Netherlands, 2016: 391–407. doi: 10.1007/978-3-319-46475-6_25.
[13]	KIM J, LEE J K, and LEE K M. Accurate image super-resolution using very deep convolutional networks[C]. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR’16), Las Vegas, USA, 2016: 1646–1654. doi: 10.1109/CVPR.2016.182.
[14]	LEDIG C, THEIS L, HUSZÁR F, et al. Photo-realistic single image super-resolution using a generative adversarial network[C]. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, USA, 2017: 105–114. doi: 10.1109/CVPR.2017.19.
[15]	LIM B, SON S H, KIM H W, et al. Enhanced deep residual networks for single image super-resolution[C]. 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Honolulu, USA, 2017: 1132–1140. doi: 10.1109/CVPRW.2017.151.
[16]	FENG Hongxiao, HOU Biao, and GONG Maoguo. SAR Image despeckling based on local homogeneous-region segmentation by using pixel-relativity measurement[J]. IEEE Transactions on Geoscience and Remote Sensing, 2011, 49(7): 2724–2737. doi: 10.1109/TGRS.2011.2107915
[17]	ZEILER M D and FERGUS R. Visualizing and understanding convolutional networks[C]. The 13th European Conference on Computer Vision (ECCV), Zurich, Switzerland, 2014: 818–833. doi: 10.1007/978-3-319-10590-1_53.
[18]	郭玥秀, 杨伟, 刘琦, 等. 残差网络研究综述[J/OL]. 计算机应用研究: 1–8. http://kreader.cnki.net/Kreader/CatalogViewPage.aspx?dbCode=CJFQ&filename=JSYJ20190606001&tablename=CAPJLAST&compose=&first=1&uid=, 2019. GUO Yuexiu, YANG Wei, LIU Qi, et al. Survey of residual network[J/OL]. Application Research of Computers: 1–8. http://kreader.cnki.net/Kreader/CatalogViewPage.aspx?dbCode=CJFQ&filename=JSYJ20190606001&tablename=CAPJLAST&compose=&first=1&uid=, 2019.
[19]	HE Kaiming, ZHANG Xingyu, REN Shaoqing, et al. Delving deep into rectifiers: Surpassing human-level performance on imagenet classification[C]. 2015 IEEE International Conference on Computer Vision (ICCV), Santiago, Chile, 2015: 1026–1034. doi: 10.1109/ICCV.2015.123.

施引文献

期刊类型引用(13)

1.	李铭典，肖顺平，陈思伟. 极化雷达图像目标超分辨率重建研究进展. 电子与信息学报. 2024(05): 1806-1826 . 百度学术
2.	向俞明，滕飞，王林徽，焦念刚，王峰，尤红建. 基于快速异源配准的高分辨率SAR影像海岛区域正射校正. 雷达学报. 2024(04): 866-884 . 本站查看
3.	康健，王智睿，祝若鑫，孙显. 基于监督对比学习正则化的高分辨率SAR图像建筑物提取方法. 雷达学报. 2022(01): 157-167 . 本站查看
4.	李宁，郭志顺，毋琳，赵建辉. River-Net：面向河道提取的Refined-Lee Kernel深度神经网络模型. 雷达学报. 2022(03): 324-334 . 本站查看
5.	郭拯危，付泽文，李宁，白澜 . 高分辨率斜视聚束SAR回波仿真加速算法研究. 计算机科学. 2022(08): 178-183 . 百度学术
6.	陈嘉琪，刘祥梅，李宁，张燕. 一种超分辨SAR图像水域分割算法及其应用. 电子与信息学报. 2021(03): 700-707 . 百度学术
7.	曲延云，陈蓉，李翠华，王菡子. 深度学习单帧图像超分辨率重建研究综述. 厦门大学学报(自然科学版). 2021(03): 555-570+641-642 . 百度学术
8.	李闯，赵建辉，戚雯雯，常继科. 多源遥感水库蓄水量估算系统. 计算机系统应用. 2021(05): 66-75 . 百度学术
9.	李宁，吕宗森，郭拯危. 联合变化检测与子带对消技术的SAR图像干扰抑制方法. 系统工程与电子技术. 2021(09): 2484-2492 . 百度学术
10.	敬绍迪，喻玲娟，胡跃虹，杨泽洲，卢忠亮，谢晓春. 基于U-Net和胶囊网络的合成孔径雷达图像语义分割. 激光与光电子学进展. 2021(20): 156-165 . 百度学术
11.	马彦恒，侯建强，张炜民，李根. 基于MN-MEA算法的无人机载SAR相位误差补偿处理. 系统工程与电子技术. 2020(09): 1945-1952 . 百度学术
12.	戴牧宸，冷祥光，熊博莅，计科峰. 基于改进双边网络的SAR图像海陆分割方法. 雷达学报. 2020(05): 886-897 . 本站查看
13.	毋琳，牛世林，郭拯危，李宁. 一种新颖的亚像素级SAR图像水陆分割方法. 雷达科学与技术. 2020(05): 494-500+508 . 百度学术