基于局部拟合一致性的机载SAR图像与仿真SAR图像匹配方法

陈彦铭; 张帆; 何岷; 麻丽香; 汪丙南

doi:10.12000/JR26027

基于局部拟合一致性的机载SAR图像与仿真SAR图像匹配方法

DOI: 10.12000/JR26027 CSTR: 32380.14.JR26027

陈彦铭¹,
张帆¹,
何岷^2, ,,
麻丽香³,
汪丙南⁴

1.
北京化工大学信息科学与技术学院北京 100029
2.
北京遥感设备研究所北京 100854
3.
中国空间技术研究院北京 100094
4.
中国科学院空天信息创新研究院北京 100190

基金项目: 微波成像全国重点实验室开放基金

详细信息

作者简介:
陈彦铭，博士生，主要研究方向为SAR, InSAR系统理论及应用等

张　帆，教授，主要研究方向为SAR系统仿真、SAR图像解译、人工智能等

何　岷，研究员，主要研究方向为雷达总体技术

麻丽香，研究员，主要研究方向为干涉SAR信号处理及应用等

汪丙南，研究员，主要研究方向为干涉SAR理论应用、数字阵列信号处理等

通讯作者:
何岷 hemindreams@sina.com

责任主编：朱岱寅 Corresponding Editor: ZHU Daiyin

中图分类号: TN957.52
计量
- 文章访问数:
- HTML全文浏览量:
- PDF下载量:
- 被引次数: 0
出版历程
- 收稿日期: 2026-01-21
- 修回日期: 2026-03-12

Matching Method for Airborne and Simulated Synthetic Aperture Radar Images Based on Local Fitting Consistency

1.
School of Information Science and Technology, Beijing University of Chemical Technology, Beijing 100029, China
2.
Beijing Institute of Remote Sensing Equipment, Beijing 100854, China
3.
China Academy of Space Technology, Beijing 100094, China
4.
Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100190, China

Funds: Open Foundation for National Key Laboratory of Microwave Imaging

More Information

Corresponding author: HE Min, hemindreams@sina.com

摘要

摘要: 成像几何差异是引起SAR图像特征相对畸变的主要原因，进而导致SAR图像的匹配难度骤增。以仿真SAR图像作为参考图可以从根本上消除因成像几何差异造成的图像特征相对畸变，然而实测SAR图像与仿真SAR图像仍然会存在散射和噪声方面的差异，且现有的匹配算法基本采用对称式关键点检测及描述符匹配，因此匹配点的数量和精度仍然有待提升。针对上述问题，该文根据实测和仿真SAR图像的局部统计特征提出了非对称式的局部拟合一致性相似度度量准则，并基于该相似度设计了机载SAR图像与仿真SAR图像的粗匹配和精匹配方法，在此基础上引入地形特征提高关键点检测的多样性，最终实现实测机载SAR图像与仿真SAR图像的鲁棒匹配。实验结果表明，在不同噪声程度下，基于局部拟合一致性度量准则设计匹配方法具备更强的鲁棒性和准确性，在匹配精度等多方面指标上均显著优于现有主流算法。
- 合成孔径雷达 /
- 后向散射系数 /
- SAR仿真 /
- 图像匹配 /
- 局部拟合一致性
Abstract: Variations in imaging geometry are the main cause of relative feature distortion in Synthetic Aperture Radar (SAR) images, greatly increasing the difficulty of image matching. Using simulated SAR images as references can remove the feature distortions caused by geometric differences. However, significant differences in scattering characteristics and noise patterns between measured and simulated images still exist. Additionally, since most existing matching algorithms mainly rely on symmetric keypoint detection and descriptor matching, the number and precision of matched points are not optimal. To solve these problems, this paper introduces an asymmetric Local Fitting Consistency (LFC) similarity metric based on the local statistical features of both measured and simulated SAR images. Using this metric, a coarse-to-fine matching framework for airborne and simulated SAR images is designed. Furthermore, terrain features are added to improve keypoint detection diversity, leading to more robust matching between airborne and simulated SAR images. Experimental results show that the proposed LFC-based matching method offers better robustness and accuracy compared to other approaches, significantly surpassing current state-of-the-art algorithms in terms of matching precision and other key metrics.
- Synthetic Aperture Radar (SAR) /
- Backscattering coefficient /
- SAR simulation /
- Image matching /
- Local fitting consistency.

HTML全文

图 1 机载SAR成像几何

Figure 1. Airborne SAR imaging geometry

下载: 全尺寸图片幻灯片

图 2 真实SAR图像与通过不同模型仿真生成的参考图对比

Figure 2. Comparison of real SAR images with reference images simulated by different models

下载: 全尺寸图片幻灯片

图 3 关键点检测和匹配流程

Figure 3. Keypoint detection and matching pipeline

下载: 全尺寸图片幻灯片

图 4 地形特征提取及关键点检测

Figure 4. Terrain Feature Extraction and Keypoint Detection

下载: 全尺寸图片幻灯片

图 5 地形关键点扩散示意图

Figure 5. Schematic diagram of terrain keypoint diffusion

下载: 全尺寸图片幻灯片

图 6 UAVSAR机载实测SAR图像

Figure 6. UAVSAR airborne measured SAR image

下载: 全尺寸图片幻灯片

图 7 粗匹配实验区域DEM

Figure 7. DEM of the coarse matching experimental area

下载: 全尺寸图片幻灯片

图 8 精匹配对照组哨兵一号跨视角星载SAR图像

Figure 8. Sentinel-1 cross-view spaceborne SAR image for the fine matching control group

下载: 全尺寸图片幻灯片

图 9 模板匹配实验结果热力图

Figure 9. Heatmap of template matching experiment results

下载: 全尺寸图片幻灯片

图 10 实测机载SAR图像与星载SAR及仿真参考图特征匹配结果

Figure 10. Feature matching results of measured airborne SAR image with spaceborne SAR and simulated reference images

下载: 全尺寸图片幻灯片

1 地形特征点检测方法

1. Terrain feature point detection method

输入：数字高程模型矩阵D
输出：特征点二值掩模矩阵K
(1) 山脊线/山谷线骨架提取
$ ({\boldsymbol{R}}_{\text{skel}},{\boldsymbol{V}}_{\text{skel}},{\boldsymbol{R}}_{\text{raw}},{\boldsymbol{V}}_{\text{raw}},{\boldsymbol{R}}_{\text{dil}},{\boldsymbol{V}}_{\text{dil}}) $ 　　$\leftarrow \text{DetectRidgesValleys}(\boldsymbol{D}) $
其中$ {\boldsymbol{R}}_{\text{skel}} $与$ {\boldsymbol{V}}_{\text{skel}} $分别为骨化后的山脊线与山谷线二值矩阵。
(2) 骨架分支修剪
设定最小分支长度阈值L_min = 15，分别处理山脊与山谷骨架：
$ {\boldsymbol{R}}_{\mathrm{skel}}^{\mathrm{proc}}\leftarrow \text{ProcessSkeletons}({\boldsymbol{R}}_{\mathrm{skel}},{L}_{\min }) $，　　　 $ {\boldsymbol{V}}_{\mathrm{skel}}^{\mathrm{proc}}\leftarrow \text{ProcessSkeletons}({\boldsymbol{V}}_{\mathrm{skel}},{L}_{\min }) $
(3) 分支点提取
分别从处理后的骨架中提取分支点：
$ {\boldsymbol{B}}_{R}\leftarrow \text{ExtractBranchPoints}(\boldsymbol{R}_{\mathrm{skel}}^{\mathrm{proc}}) $，　　　 $ {\boldsymbol{B}}_{V}\leftarrow \text{ExtractBranchPoints}(\boldsymbol{V}_{\mathrm{skel}}^{\mathrm{proc}}) $
(4) 特征点融合
初始化与D尺寸相同的二值矩阵K，并将分支点位置置为真
$ {\boldsymbol{K}}[{\boldsymbol{B}}_{R}]\leftarrow \text{True}, {\boldsymbol{K}}[{\boldsymbol{B}}_{V}]\leftarrow \text{True} $
(5) 返回K

下载: 导出CSV

表 1 SAR图像粗匹配实验各方法主要参数设置情况

Table 1. Parameter settings of main methods in SAR image coarse matching experiment

方法	主要参数及预设值	参考来源
MIND	patch size = 3×3×3; search region: 6-neighbourhood	文献[24]
CFOG	Gaussian STD = 0.8; channel number m = 9	文献[12]
HOPES	k_MSG = 1.4, N_Scale = 4, Orientation M = 8	文献[25]
PSOC	Fast-MSG = 3, σ₁=2, σ_i / σ_{i – 1}=1.6	文献[26]
MIRD	Gaussian Kernel Size: 9×9; k = 0.9	文献[11]
OMIRD	Gaussian Kernel Size: 3×3	文献[27]
LFC	r = 32	—

下载: 导出CSV

表 2 不同噪声条件下各模板匹配方法热力图峰值x和y方向偏离结果(pixel)

Table 2. Deviation results of heatmap peaks in x and y directions for different template matching methods under various noise conditions (Pixel)

模板匹配方法	噪声方差
模板匹配方法	0	0.1	0.2
NCC	(+3,+3)	(+3,+3)	(+3,+3)
MIND	(+3,+3)	(+3,+3)	(+3,+4)
CFOG	(–2,–2)	(–2,–3)	(–2,–3)
HOPES	(+2,+2)	(+2,+3)	(–126,+3)
PSOC	(+12,–13)	(+12,-13)	(+107,-18)
OMIRD	(+3,+3)	(+3,+3)	(+3,+3)
MIRD	(+3,+3)	(–127,–21)	(–127,–21)
LFC	(+1,+2)	(+2,+2)	(+2,+2)

下载: 导出CSV

表 3 不同噪声条件下各模板匹配方法匹配误差(pixel)

Table 3. Matching errors of different template matching methods under various noise conditions (Pixel)

模板匹配方法	噪声方差
模板匹配方法	0	0.1	0.2
NCC	16.97	16.97	16.97
MIND	16.97	16.97	20
CFOG	11.31	14.42	14.42
HOPES	11.31	14.42	504.14
PSOC	70.77	70.77	434.01
OMIRD	16.97	16.97	16.97
MIRD	16.97	514.9	514.9
LFC	8.94	11.31	11.31

下载: 导出CSV

表 4 不同噪声条件下各模板匹配方法对应热力图峰值梯度对比结果

Table 4. Comparison of heatmap peak gradients for different template matching methods under various noise conditions

模板匹配方法	噪声方差
模板匹配方法	0	0.1	0.2
NCC	106.24	106.17	106.59
MIND	130.12	120.26	100.96
CFOG	58.29	58.73	58.45
HOPES	67.68	56.84	10.33(×)
PSOC	174.2(×)	186.4(×)	246.57(×)
OMIRD	114.53	85.53	80.53
MIRD	143.44	143.03(×)	141.47(×)
LFC	629.36	321.22	226.07
注：由于HOPES, PSOC, MIRD的部分匹配误差过大，严重偏离目标点，因此我们认为其热力图峰值已经不具备参考性，在表中以“×”符号标识。

下载: 导出CSV

表 5 SAR图像精匹配实验各方法主要参数设置情况

Table 5. Parameter settings of main methods in sar image fine matching experiment

方法	主要参数及预设值	参考来源
KAZE-SAR	default	文献[6]
SAR-SIFT	DistRatio = 0.9; LayerNum = 8	文献[5]
ASS	Orientations = 8; Radial Bins Num = 3; Region Radius = 42	文献[29]
FED-HOPC	Template Size = 100; searchRad = 10; gamma1 = 0.15	文献[30]
M2FF	Filter Orientations = 4; Filter Scales = 4	文献[28]
LFC / S_joint	r = 32, $ {w}_{1}=30 $, $ {w}_{2}=64 $	—

下载: 导出CSV

表 6 不同噪声条件下各匹配方法匹配确率等评估

Table 6. Evaluation of matching accuracy for different methods under various noise conditions

方法	评价指标
	RMSE(像素)			TNMs(个)			NCMs(个)			CMR(%)
	σ = 0	σ = 0.1	σ = 0.2	σ = 0	σ = 0.1	σ = 0.2	σ = 0	σ = 0.1	σ = 0.2	σ = 0	σ = 0.1	σ = 0.2
KAZE-SAR	1.94	1.96	1.93	806	629	278	397	294	141	0.493	0.467	0.507
SAR-SIFT	1.57	3.21	3.94	122	34	14	86	8	2	0.705	0.235	0.143
ASS	1.63	2.12	1.98	590	333	291	416	180	141	0.705	0.541	0.485
FED-HOPC	2.93	5.35	10.15	100	100	100	36	23	6	0.36	0.23	0.06
M2FF	1.41	1.6	9.27	317	213	27	239	156	11	0.754	0.732	0.407
本文	1.39	1.39	1.48	1130	1334	925	1017	1213	809	0.9	0.909	0.875

下载: 导出CSV

参考文献(30)

[1]	靳国旺, 张红敏, 徐青. 雷达摄影测量[M]. 北京: 测绘出版社, 2015. JIN Guowang, ZHANG Hongmin, and XU Qing. Radargrammetry[M]. Beijing: Surveying and Mapping Press, 2015.
[2]	孙晓坤, 贠泽楷, 胡粲彬, 等. 面向高分辨率多视角SAR图像的端到端配准算法[J]. 雷达学报, 2025, 14(2): 389–404. doi: 10.12000/JR24211. SUN Xiaokun, YUN Zekai, HU Canbin, et al. End-to-end registration algorithm for high-resolution multi-view SAR images[J]. Journal of Radars, 2025, 14(2): 389–404. doi: 10.12000/JR24211.
[3]	LOWE D G. Distinctive image features from scale-invariant keypoints[J]. International Journal of Computer Vision, 2004, 60(2): 91–110. doi: 10.1023/B:VISI.0000029664.99615.94.
[4]	ALCANTARILLA P F, BARTOLI A, and DAVISON A J. KAZE features[C]. –The 12th European Conference on Computer Vision, Florence, Italy, 2012: 214–227. doi: 10.1007/978-3-642-33783-3_16.
[5]	DELLINGER F, DELON J, GOUSSEAU Y, et al. SAR-SIFT: A SIFT-like algorithm for applications on SAR images[C]. 2012 IEEE International Geoscience and Remote Sensing Symposium, Munich, Germany, 2012: 3478–3481. doi: 10.1109/IGARSS.2012.6350671.
[6]	POURFARD M, HOSSEINIAN T, SAEIDI R, et al. KAZE-SAR: SAR image registration using KAZE detector and modified SURF descriptor for tackling speckle noise[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 5207612. doi: 10.1109/TGRS.2021.3084411.
[7]	WANG Linhui, XIANG Yuming, YOU Hongjian, et al. A robust multiscale edge detection method for accurate SAR image registration[J]. IEEE Geoscience and Remote Sensing Letters, 2023, 20: 4006305. doi: 10.1109/LGRS.2023.3279141.
[8]	SUN Jianjun, ZHAO Yan, LI Xinbo, et al. Fractional order spectrum in SAR image registration[C]. 2024 IEEE International Conference on Multimedia and Expo (ICME), Niagara Falls, Canada, 2024: 1–6. doi: 10.1109/ICME57554.2024.10688291.
[9]	XIANG Deliang, PAN Xiaoyu, DING Huaiyue, et al. Two-stage registration of SAR images with large distortion based on superpixel segmentation[J]. IEEE Transactions on Geoscience and Remote Sensing, 2024, 62: 5211115. doi: 10.1109/TGRS.2024.3392971.
[10]	YE Yuanxin, SHAN Jie, BRUZZONE L, et al. Robust registration of multimodal remote sensing images based on structural similarity[J]. IEEE Transactions on Geoscience and Remote Sensing, 2017, 55(5): 2941–2958. doi: 10.1109/TGRS.2017.2656380.
[11]	LIU Xuecong, TENG Xichao, LUO Jing, et al. Robust multi-sensor image matching based on normalized self-similarity region descriptor[J]. Chinese Journal of Aeronautics, 2024, 37(1): 271–286. doi: 10.1016/j.cja.2023.10.003.
[12]	YE Yuanxin, BRUZZONE L, SHAN Jie, et al. Fast and robust matching for multimodal remote sensing image registration[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(11): 9059–9070. doi: 10.1109/TGRS.2019.2924684.
[13]	项德良, 徐益豪, 程建达, 等. 一种基于特征交汇关键点检测和Sim-CSPNet的SAR图像配准算法[J]. 雷达学报, 2022, 11(6): 1081–1097. doi: 10.12000/JR22110. XIANG Deliang, XU Yihao, CHENG Jianda, et al. An algorithm based on a feature interaction-based keypoint detector and Sim-CSPNet for SAR image registration[J]. Journal of Radars, 2022, 11(6): 1081–1097. doi: 10.12000/JR22110.
[14]	ZHAO Xuanran, WU Yan, HU Xin, et al. A novel dual-branch global and local feature extraction network for SAR and optical image registration[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024, 17: 17637–17650. doi: 10.1109/JSTARS.2024.3435684.
[15]	WU Wei, XIAN Yong, SU Juan, et al. A Siamese template matching method for SAR and optical image[J]. IEEE Geoscience and Remote Sensing Letters, 2022, 19: 4017905. doi: 10.1109/LGRS.2021.3108579.
[16]	HU Xin, WU Yan, LIU Xingyu, et al. Intra- and inter-modal graph attention network and contrastive learning for SAR and optical image registration[J]. IEEE Transactions on Geoscience and Remote Sensing, 2023, 61: 5220216. doi: 10.1109/TGRS.2023.3328368.
[17]	TOSS T, DÄMMERT P, SJANIC Z, et al. Navigation with SAR and 3D-map aiding[C]. 2015 18th International Conference on Information Fusion (Fusion), Washington, USA, 2015: 1505–1510.
[18]	李贺, 秦志远, 靳国旺. 基于DEM的SAR图像模拟[J]. 测绘科学技术学报, 2008, 25(4): 296–299. doi: 10.16251/j.cnki.1009-2307.2013.04.040. LI He, QIN Zhiyuan, and JIN Guowang. Simulation of SAR images based on DEM[J]. Journal of Geomatics Science and Technology, 2008, 25(4): 296–299. doi: 10.16251/j.cnki.1009-2307.2013.04.040.
[19]	张红敏, 靳国旺, 徐青, 等. 机载SAR图像与仿真SAR图像的匹配策略[J]. 测绘科学技术学报, 2013, 30(2): 144–148. doi: 10.3969/j.issn.1673-6338.2013.02.009. ZHANG Hongmin, JIN Guowang, XU Qing, et al. Matching strategy of airborne SAR image and simulated SAR image[J]. Journal of Geomatics Science and Technology, 2013, 30(2): 144–148. doi: 10.3969/j.issn.1673-6338.2013.02.009.
[20]	HE Kaiming, SUN Jian, and TANG Xiaoou. Guided image filtering[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35(6): 1397–1409. doi: 10.1109/TPAMI.2012.213.
[21]	Jet Propulsion Laboratory. UAVSAR L-band Synthetic Aperture Radar (SAR) repeat pass interferometry products[R]. NASA Alaska Satellite Facility DAAC, 2024.
[22]	European Space Agency. Copernicus global digital elevation models[R]. OT.032021.4326.1, 2021.
[23]	European Space Agency. Sentinel-1[R]. NASA Alaska Satellite Facility DAAC, 2025.
[24]	HEINRICH M P, JENKINSON M, BHUSHAN M, et al. MIND: Modality independent neighbourhood descriptor for multi-modal deformable registration[J]. Medical Image Analysis, 2012, 16(7): 1423–1435. doi: 10.1016/j.media.2012.05.008.
[25]	LI Shuo, LV Xiaolei, REN Jian, et al. A robust 3D density descriptor based on histogram of oriented primary edge structure for SAR and optical image co-registration[J]. Remote Sensing, 2022, 14(3): 630. doi: 10.3390/rs14030630.
[26]	LI Shuo, LV Xiaolei, WANG Hao, et al. A novel fast and robust multimodal images matching method based on primary structure-weighted orientation consistency[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2023, 16: 9916–9927. doi: 10.1109/JSTARS.2023.3325577.
[27]	TENG Xichao, LIU Xuecong, LI Zhang, et al. OMIRD: Orientated modality independent region descriptor for optical-to-SAR image matching[J]. IEEE Geoscience and Remote Sensing Letters, 2023, 20: 4003405. doi: 10.1109/LGRS.2023.3256186.
[28]	XIONG Xin, JIN Guowang, XU Qing, et al. Robust registration algorithm for optical and SAR images based on adjacent self-similarity feature[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 5233117. doi: 10.1109/tgrs.2022.3197357.
[29]	YE Yibin, WANG Qinwei, ZHAO Hong, et al. Fast and robust optical-to-SAR remote sensing image registration using region-aware phase descriptor[J]. IEEE Transactions on Geoscience and Remote Sensing, 2024, 62: 5208512. doi: 10.1109/TGRS.2024.3379370.
[30]	ZHANG Xiaoting, WANG Yinghua, LIU Jun, et al. Robust coarse-to-fine registration algorithm for optical and SAR images based on two novel multiscale and multidirectional features[J]. IEEE Transactions on Geoscience and Remote Sensing, 2024, 62: 5215126. doi: 10.1109/TGRS.2024.3417217.