基于动态聚合网络的SAR目标检测

doi:10.12305/j.issn.1001-506X.2025.08.11

系统工程与电子技术 ›› 2025, Vol. 47 ›› Issue (8): 2527-2539.doi: 10.12305/j.issn.1001-506X.2025.08.11

• 传感器与信号处理 • 上一篇

基于动态聚合网络的SAR目标检测

倪康¹^,²(), 贾文杰¹(), 邹旻瑞¹(), 郑志忠¹^,²^,*()

1. 南京邮电大学计算机学院、软件学院、网络空间安全学院，江苏南京 210023
2. 江苏省航空对地探测与智能感知工程研究中心，江苏南京 210049

收稿日期:2024-05-10 出版日期:2025-08-25 发布日期:2025-09-04
通讯作者: 郑志忠 E-mail:tznikang@163.com;Shadow_Armor@163.com;traveler_wood@163.com;zhengzz_js@126.com
作者简介:倪　康（1990—），男，副教授，博士，主要研究方向为SAR图像目标检测与分类
贾文杰（2001—），男，硕士研究生，主要研究方向为SAR图像目标检测
邹旻瑞（1999—），男，硕士研究生，主要研究方向为SAR图像目标检测
基金资助:
国家自然科学基金（62101280）；江苏省自然科学基金（BK20210588）；中国博士后科学基金（2023M731781）；江苏省航空对地探测与智能感知工程中心开放基金（JSECF2023-01，JSECF2023-05）；雷达成像与微波光子技术教育部重点实验室（南京航空航天大学）基金（NJ20230005）；南京邮电大学引进人才科研启动基金（NY222107）资助课题

SAR object detection based on dynamic aggregation network

Kang NI¹^,²(), Wenjie JIA¹(), Minrui ZOU¹(), Zhizhong ZHENG¹^,²^,*()

1. School of Computer Science，Nanjing University of Posts and Telecommunications，Nanjing 210023，China
2. Jiangsu Province Engineering Research Center of Airborne Detecting and Intelligent Perceptive Technology，Nanjing 210049，China

Received:2024-05-10 Online:2025-08-25 Published:2025-09-04
Contact: Zhizhong ZHENG E-mail:tznikang@163.com;Shadow_Armor@163.com;traveler_wood@163.com;zhengzz_js@126.com

摘要/Abstract

摘要：

针对合成孔径雷达（synthetic aperture radar，SAR）图像的成像机理导致SAR图像受噪声影响严重且SAR目标类别存在不均衡问题，使得SAR目标特征刻画较困难。针对上述问题，提出一种基于动态聚合网络（dynamic aggregation network，DANet）的SAR目标检测方法。该网络以无锚框全卷积一阶段目标检测（fully convolutional one-stage object detection，FCOS）网络为基础网络框架。DANet在特征金字塔网络中嵌入具有动态坐标注意力的动态聚合模块，以提高噪声影响下SAR目标特征学习能力。为了缓解SAR数据集中不同类别之间数量不平衡的问题，DANet在检测头的回归分支引入类平衡动态交并比（intersection over union，IoU）损失函数,通过动态非单调机制和类平衡因子引导模型参数更新。在MSAR-1.0数据集和SAR-AIRcraft-1.0数据集上的实验结果表明，DANet在上述SAR目标检测数据集上的目标检测准确率分别达到了70.25%和52.36%，相比基准网络FCOS，其平均精度分别提高了2.73%和3.67%。与其他相关算法相比，DANet在SAR图像目标检测精度方面具有明显优势。

关键词: 合成孔径雷达, 注意力机制, 目标检测, 特征金字塔, 损失函数

Abstract:

Due to the imaging mechanism of synthetic aperture radar (SAR) images, SAR images are severely affected by noise and there is an imbalance in SAR target categories, making it difficult to characterize SAR target features. In view of these problems, a SAR target detection method based on dynamic aggregation network (DANet) is proposed. This network is based on the fully convolutional one-stage object detection (FCOS) network without anchor boxes. DANet embeds a dynamic aggregation module with dynamic coordinate attention in the feature pyramid network to improve the SAR target feature learning ability under noise influence. In order to alleviate the problem of imbalanced numbers between different categories in SAR datasets, DANet introduces a class balanced dynamic intersection over union (IoU) loss function in the regression branch of the detection head, which guides model parameter updates through dynamic non monotonic mechanisms and class balance factors. The experimental results on the MSAR-1.0 dataset and SAR-AIRcraft-1.0 dataset show that DANet achieved target detection accuracies of 70.25% and 52.36%, respectively, on the aforementioned SAR target detection datasets. Compared with the benchmark network FCOS, its average accuracy increased by 2.73% and 3.67%, respectively. Compared with other related algorithms, DANet has significant advantages in SAR image target detection accuracy.

Key words: synthetic aperture radar (SAR), attention mechanism, object detection, feature pyramid, loss function

中图分类号:

TP 751

倪康, 贾文杰, 邹旻瑞, 郑志忠. 基于动态聚合网络的SAR目标检测[J]. 系统工程与电子技术, 2025, 47(8): 2527-2539.

Kang NI, Wenjie JIA, Minrui ZOU, Zhizhong ZHENG. SAR object detection based on dynamic aggregation network[J]. Systems Engineering and Electronics, 2025, 47(8): 2527-2539.

图/表 20

图1

图2

图3

图4

表1

表2

图5

表3

表4

图6

表5

表6

图7

图8

图9

图10

表7

表8

表9

表10

参考文献 39

1	NI J C, LUO Y, WANG D, et al. Saliencybased SAR target detection via convolutional sparse feature enhancement and bayesian inference[J]. IEEE Trans. on Geoscience and Remote Sensing, 2023, 61, 5202015.
2	ZHOU P, WANG P, CAO J, et al. PSFNet: efficient detection of SAR image based on petty-specialized feature aggregation[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2023, 17, 190- 205.
3	方小宇, 黄丽佳. 基于全局位置信息和残差特征融合的SAR船舶检测算法[J]. 系统工程与电子技术, 2024, 46 (3): 839- 848.
	FANG X Y, HUANG L J. SAR ship detection algorithm based on global position information and fusion of residual feature[J]. Systems Engineering and Electronics, 2024, 46 (3): 839- 848.
4	扈琪, 胡绍海, 刘帅奇. 基于多层显著性模型的SAR图像舰船目标检测[J]. 系统工程与电子技术, 2024, 46 (2): 478- 487.
	HU Q, HU S H, LIU S Q. Ship detection in SAR image based on multi-layer saliency model[J]. Systems Engineering and Electronics, 2024, 46 (2): 478- 487.
5	ZHANG S F, CHI C, YAO Y Q, et al. Bridging the gap between anchor-based and anchor-free detection via adaptive training sample selection[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 9759−9768.
6	ZHOU X, WANG D, KRÄHENBÜHL P, Objects as points[EB/OL]. [2024-04-10]. https://arxiv.org/abs/1904.07850.
7	REN S Q, HE K M, GIRSHICK R, et al. Faster rcnn: towards real-time object detection with region proposal networks[J]. IEEE Trans. on Pattern Analysis and Machine Intelligence, 2017, 39 (6): 1137- 1149. doi: 10.1109/TPAMI.2016.2577031
8	TIAN Z, SHEN C H, CHEN H, et al. FCOS: fully convolutional one-stage object detection[C]//Proc. of the IEEE/CVF International Conference on Computer Vision, 2019 .
9	TIAN Z, SHEN C H, CHEN H, et al. FCOS: a simple and strong anchor-free object detector[J]. IEEE Trans. on Pattern Analysis and Machine Intelligence, 2020, 44 (4): 1922- 1933.
10	YANG Z, LIU S H, HU H, et al. Reppoints: point set representation for object detection[C]//Proc. of the IEEE/CVF International Conference on Computer Vision, 2019: 9657−9666.
11	LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[C]//Proc. of the IEEE International Conference on Computer Vision, 2017: 2980−2988.
12	REDMON J, FARHADI A. Yolov3: an incremental improvement[EB/OL]. [2024-04-10]. https://arxiv.org/pdf/1804.02767.
13	CHEN Q, WANG Y M, YANG T, et al. You only look one-level feature[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021: 13039−13048.
14	LIU Y B, ZHANG F, MA L, et al. Longtailed SAR target recognition based on expert network and intra-class resampling[J]. IEEE Geoscience and Remote Sensing Letters, 2023, 20, 4010405.
15	LI Q, JIANG Z J, ZHOU X Q. Ship detection for SAR remote sensing images based oncoordinate-enhanced attention and long-short-range spatial semantic fused context[C]//Proc. of the 3rd China International SAR Symposium, 2022.
16	PU X Y, JIA H C, XU F. Cross-domain SAR ship detection in strong interference environment based on image-to-image translation[C]//Proc. of the IGARSS IEEE International Geoscience and Remote Sensing Symposium, 2023: 1798−1801.
17	ZHANG Y W, CAO Y, FENG X B, et al. SAR object detection encounters deformed complex scenes and aliased scattered power distribution[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2022, 15, 4482- 4495. doi: 10.1109/JSTARS.2022.3157749
18	WANG C, RUAN R, ZHAO Z C, et al. Category-oriented localization distillation for SAR object detection and a unified benchmark[J]. IEEE Trans. on Geoscience and Remote Sensing, 2023, 61, 5211314.
19	SUN G C, XING M, XIA X G, et al. A unified focusing algorithm for several modes of SAR based on FrFT[J]. IEEE Trans. on Geoscience and Remote Sensing, 2012, 51 (5): 3139- 3155.
20	XIE Z F, WANG X Q, CHEN B. SAR ship recognition: dataset construction and long tail classification method[C]//Proc. of the 3rd International Conference on Electronics, Communications and Information Technology, 2022: 88−93.
21	LIN T Y, DOLLÁR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]//Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2017: 2117−2125.
22	陈杰, 黄志祥, 夏润繁, 等. 大规模多类SAR目标检测数据集-1.0[EB/OL]. [2024-04-10]. https://radars.ac.cn/web/data/getData?dataType=MSAR.
	CHEN J, HUANG Z X, XIA R F, et al. Large-scale multi-class SAR image target detection dataset-1.0[EB/OL]. [2024-04-10]. https://radars.ac.cn/web/data/getData?dataType=MSAR.
23	王智睿, 康玉卓, 曾璇, 等. SAR-AIRcraft-1.0: 高分辨率 SAR 飞机检测识别数据集[J]. 雷达学报, 2023, 12 (4): 906- 922. doi: 10.12000/JR23043
	WANG Z R, KANG Y Z, ZENG X, et al. SAR-AIRcraft-1.0: high-resolution SAR aircraft detection and recognition dataset[J]. Journal of Radars, 2023, 12 (4): 906- 922. doi: 10.12000/JR23043
24	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 770−778.
25	CUI Y, JIA M, LIN T Y, et al. Class-balanced loss based on effective number of samples[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 9268−9277.
26	HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]//Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 7132−7141.
27	PARK J, WOO S, LEE J Y, et al. Bam: bottleneck attention module[EB/OL]. [2024-04-10]. https://arxiv.org/abs/1807.06514v1.
28	FU J, LIU J, TIAN H J, et al. Dual attention network for scene segmentation[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 3146−3154.
29	WOO S, PARK J, LEE J Y, et al. Cbam: convolutional block attention module[C]//Proc. of the European Conference on Computer Vision, 2018: 3−19.
30	HOU Q B, ZHOU D Q, FENG J S. Coordinate attention for efficient mobile network design[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021: 13713−13722.
31	WU P, HUANG H H, QIAN H X, et al. SRCANet: stacked residual coordinate attention network for infrared ship detection[J]. IEEE Trans. on Geoscience and Remote Sensing, 2022, 60, 5003614.
32	HOWARD A, SANDLER M, CHU G, et al. Searching for mobilenetv3[C]//Proc. of the IEEE/CVF International Conference on Computer vision, 2019: 1314−1324.
33	TONG Z J, CHEN Y H, XU Z W, et al. Wise-IoU: bounding box regression loss with dynamic focusing mechanism[EB/OL]. [2024-04-10]. https://arxiv.org/abs/2301.10051.
34	EVERINGHAM M, VAN GOOL L, WILLIAMS CK I, et al. The pascal visual object classes （VOC） challenge[J]. International Journal of Computer Vision, 2010, 88, 303- 338. doi: 10.1007/s11263-009-0275-4
35	CHEN K, WANG J Q, PANG J M, et al. MMDetection: open MMlab detection toolbox and benchmark[EB/OL]. [2024-04-10]. https://arxiv.org/abs/1906.07155.
36	ZHU X Z, SU W J, LU L W, et al. Deformable DETR: deformable transformers for end-to-end object detection[EB/OL]. [2024-04-10]. https://arxiv.org/abs/2010.04159.
37	LI X, WANG W H, WU L J, et al. Generalized focal loss: learning qualified and distributed bounding boxes for dense object detection[J]. Advances in Neural Information Processing Systems, 2020, 33, 21002- 21012.
38	KIM K, LEE H S. Probabilistic anchor assignment with IoU prediction for object detection[C]//Proc. of the 16th European Conference on Computer Vision, 2020: 355−371.
39	姚群力, 胡显, 雷宏. 基于多尺度融合特征卷积神经网络的遥感图像飞机目标检测[J]. 测绘学报, 2019, 48 (10): 1266- 1274. doi: 10.11947/j.AGCS.2019.20180398
	YAO Q L, HU X, LEI H. Remote sensing image aircraft target detection based on multiscale fusion featureconvolutional neural network[J]. Acta Geodaetica et Cartographica Sinica, 2019, 48 (10): 1266- 1274. doi: 10.11947/j.AGCS.2019.20180398

类别	训练集	验证集	合计
船只	27459	12399	39858
油罐	8153	4166	12319
桥梁	1284	567	1851
飞机	5006	1362	6368
合计	41902	18494	6368

类别	训练集	验证集	测试集	合计
A330	251	27	31	309
A320/321	1637	82	52	1771
A220	3023	247	460	3730
ARJ21	447	378	362	1187
Boeing737	1755	252	550	2557
Boeing787	1930	261	454	2645
Other	2508	715	1041	4246
合计	11551	1962	2950	16363

对比算法	mAP/%	DR/%	FAR/%	FLOPS（×10⁸）	FPS	参数量（×10⁶）
ATSS^[5]	68.00	71.43	35.22	12.89	180.72	32.12
Deformable_DETR^[36]	65.02	73.51	43.57	14.62	107.61	40.10
Faster R-CNN^[7]	55.69	57.83	47.09	26.13	138.24	41.36
FCOS^[8−9]	67.52	72.10	35.27	12.89	195.97	32.12
GFL^[37]	67.24	70.75	37.25	13.08	178.14	32.27
PAA^[38]	63.37	68.73	42.10	12.89	106.52	32.12
RepPoints^[10]	49.25	56.05	57.21	12.12	106.99	36.82
RetinaNet^[11]	33.71	52.28	76.23	13.14	186.30	36.39
YOLOF^[13]	46.39	50.75	59.55	6.58	287.21	42.41
YOLOv8	58.86	58.79	36.21	8.10	303.41	30.06
DANet（本文方法）	70.25	73.23	33.50	12.94	179.41	33.25

对比算法	mAP/%	DR/%	FAR/%	FLOPS（×10⁸）	FPS	参数量（×10⁶）
ATSS^[5]	46.89	66.06	75.10	206.32	37.24	32.13
Deformable_DETR^[36]	43.53	83.31	94.21	200.40	25.43	40.10
Faster R-CNN^[7]	48.49	64.87	70.09	211.05	36.75	41.38
FCOS^[8−9]	48.69	72.34	73.29	206.32	38.36	32.13
GFL^[37]	45.49	65.30	76.10	209.49	37.18	32.27
PAA^[38]	47.67	80.03	88.12	206.32	16.01	32.13
RepPoints^[10]	51.11	77.89	74.56	193.98	30.89	36.82
RetinaNet^[11]	46.09	73.41	81.12	211.53	38.62	36.45
YOLOF^[13]	48.86	75.27	79.41	105.30	59.51	42.48
YOLOv8	49.73	68.02	70.34	140.23	62.78	30.07
DANet（本文方法）	52.36	73.93	69.51	206.54	37.42	33.26

类别	目标数量	FCOS			DANet
类别	目标数量	AP	DR	FAR	AP	DR	FAR
船只	12399	89.17	91.30	11.25	89.31	91.40	11.13
油罐	5935	63.28	68.91	38.05	63.28	68.82	37.59
桥梁	567	73.92	80.82	40.01	79.01	82.67	37.55
飞机	1363	43.72	47.79	50.77	49.41	50.53	48.02

基于动态聚合网络的SAR目标检测

SAR object detection based on dynamic aggregation network

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 20

参考文献 39

相关文章 15

编辑推荐

Metrics

本文评价

$\sigma $	MSAR-1.0	SAR-AIRcraft-1.0
0.95	67.77	49.80
0.995	67.89	49.56
0.9995	68.40	51.47
0.99995	67.87	50.16

$\delta $	$\alpha $	MSAR-1.0	SAR-AIRcraft-1.0
1.4	5	67.65	51.01
1.6	4	68.40	51.47
1.9	3	68.82	51.40
2.5	2	68.72	50.61

动态聚合模块	类平衡动态IoU损失	mAP	DR	FAR
×	×	67.52	72.10	35.27
√	×	69.30	73.08	34.79
×	√	68.82	73.11	35.48
√	√	70.25	73.23	33.50

[1]	蓝舒尧, 李宇, 张春华, 迟骋, 陈春辉. 垂直阵分集束控影区探测信号发射策略[J]. 系统工程与电子技术, 2025, 47(8): 2454-2462.
[2]	于营, 王春平, 徐金辉, 吕述杭, 付强, 陈明. PE-Net：一种优化剪枝的实时山体滑坡检测网络[J]. 系统工程与电子技术, 2025, 47(8): 2475-2485.
[3]	罗颖聪, 张磊, 魏少鹏, 孟智超. 联合失配滤波器的近区低旁瓣混沌波形设计算法[J]. 系统工程与电子技术, 2025, 47(8): 2511-2518.
[4]	付卫红, 彭文洪, 刘乃安. 混合注意力优化的SAR图像小目标检测方法[J]. 系统工程与电子技术, 2025, 47(8): 2519-2526.
[5]	符小卫, 王辛夷, 乔哲. 基于APIQ算法的多无人机攻防对抗策略[J]. 系统工程与电子技术, 2025, 47(7): 2205-2215.
[6]	黄丰卓, 冯东, 华洋晟, 黄晓涛. 一种改进的全息SAR相位梯度自聚焦算法[J]. 系统工程与电子技术, 2025, 47(6): 1786-1795.
[7]	张新征, 闫梦可, 朱晓林. 噪声伪标签容忍的半监督SAR目标识别[J]. 系统工程与电子技术, 2025, 47(6): 1796-1805.
[8]	贾竣皓, 陈学斌, 李璋峰, 叶春茂. 空间群目标ISAR成像与检测方法[J]. 系统工程与电子技术, 2025, 47(6): 1843-1854.
[9]	符小卫, 王辛夷, 乔哲. 基于ASDDPG算法的多无人机对抗策略[J]. 系统工程与电子技术, 2025, 47(6): 1867-1879.
[10]	刘宁, 李芳芳, 李新武, 洪文. 结合足迹和相位信息的SAR高层建筑三维重建[J]. 系统工程与电子技术, 2025, 47(5): 1469-1486.
[11]	韩静雯, 杨勇, 连静, 王威. 大掠射角下幅相极化特征融合雷达导引头舰船检测方法[J]. 系统工程与电子技术, 2025, 47(4): 1127-1135.
[12]	何肖阳, 陈小龙, 杜晓林, 苏宁远, 袁旺, 关键. 基于CBAM-Swin-Transformer迁移学习的海上微动目标分类方法[J]. 系统工程与电子技术, 2025, 47(4): 1155-1167.
[13]	贾钰嘉, 张思乾, 唐涛, 匡纲要. 强化散射特征的机载SAR实传图像盲超分辨重建[J]. 系统工程与电子技术, 2025, 47(3): 753-767.
[14]	杨志伟, 杨安东, 梁庚辰, 李相海, 李晓蕊, 刘杰. 多子带融合解动目标径向速度模糊方法[J]. 系统工程与电子技术, 2025, 47(3): 788-796.
[15]	陈世龙, 刘霖, 王晓蓓, 曾翰森, 刘亚波, 刘翔. 脉间捷变频SAR快速补偿频域成像处理算法[J]. 系统工程与电子技术, 2025, 47(3): 797-806.

类别	目标数量	FCOS			DANet
类别	目标数量	AP	DR	FAR	AP	DR	FAR
A330	31	73.52	83.91	28.43	82.63	90.33	26.38
A320/321	52	48.81	78.83	86.06	60.11	88.51	69.63
A220	450	50.48	83.92	81.82	54.21	82.22	73.95
ARJ21	362	50.02	61.58	61.60	47.28	56.91	64.69
Boeing737	550	31.90	61.62	94.34	31.84	60.03	93.02
Boeing787	454	36.04	60.79	83.05	37.70	65.24	84.72
其他	1041	50.10	75.79	77.73	52.76	74.43	74.18