基于特征融合和定位增强的SAR图像舰船目标检测方法

doi:10.12305/j.issn.1001-506X.2025.11.08

摘要/Abstract

摘要：

针对星载合成孔径雷达（synthetic aperture radar，SAR）图像舰船目标尺度不一致且易受背景杂波噪声干扰，使得SAR图像舰船目标检测结果存在准确率和定位精度较低的问题，提出一种基于特征融合和定位增强的舰船目标检测算法。首先，通过挖掘特征图相邻键之间的上下文信息，提高网络对舰船目标上下文特征的提取能力。其次，采用路径聚合特征金字塔网络缩减特征传递路径长度，用于传递高分辨定位特征信息，可提高目标边界框定位精度。最后，对全卷积检测器和Transformer检测器的检测结果进行融合，缓解了SAR图像中正负样本比例失衡的问题，并提升了中小型舰船目标的检测效果。基于实测数据集的实验结果表明，所提方法对复杂背景中的舰船目标具有较好的检测性能和泛化能力。

关键词: 合成孔径雷达, 舰船目标检测, 注意力机制, 特征融合, 目标定位

Abstract:

Due to the inconsistent scale of ship targets in spaceborne synthetic aperture radar （SAR） images and their susceptibility to background clutter noise, the accuracy and localization precision of SAR image ship target detection results are often limited. In view of these problems, a ship target detection algorithm based on feature fusion and localization enhancement is proposed. Firstly, by mining the contextual information between adjacent keys in the feature map, the ability of network to extract the contextual features of ship targets is improved. Secondly, the path aggregation feature pyramid network is adopted to shorten the length of feature transmission paths, which is used to transfer high-resolution positioning feature information, thereby improving the localization accuracy of the target bounding box. Finally, the detection results of the fully convolutional detector and the Transformer detector are fused to alleviate the problem of imbalanced positive and negative sample ratios in SAR images and improve the detection effect of small and medium-sized ship targets. Experimental results based on the measured dataset show that the proposed method has good detection performance and generalization ability in complex backgrounds for ship target.

Key words: synthetic aperture radar (SAR), ship target detection, attention mechanism, feature fusion, target location

中图分类号:

TP 751

王勇, 张博雅. 基于特征融合和定位增强的SAR图像舰船目标检测方法[J]. 系统工程与电子技术, 2025, 47(11): 3586-3597.

Yong WANG, Boya ZHANG. Ship target detection method in SAR images based on feature fusion and location enhancement[J]. Systems Engineering and Electronics, 2025, 47(11): 3586-3597.

图/表 20

图1

图2

表1

图3

图4

图5

图6

图7

图8

表2

图9

图10

图11

表3

图12

表4

图13

表5

图14

表6

参考文献 28

1	CAO R, WANG Y, LIN Y C, et al. An efficient preprocessing approach for airborne hybrid SAR and ISAR imaging of ship target based on kernel distribution[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2022, 15, 5147- 5162. doi: 10.1109/JSTARS.2022.3183196
2	薛峰涛, 孙天宇, 杨益民, 等. 基于全局特征融合的SAR图像旋转舰船目标检测算法[J]. 系统工程与电子技术, 2024, 46 (12): 4044- 4053.
	XUE F T, SUN T Y, YANG Y M, et al. Rotated ship target detection algorithm in SAR images based on global feature fusion[J]. Systems Engineering and Electronics, 2024, 46 (12): 4044- 4053.
3	QIAN G Z, WANG Y, MAO J. Estimation of Doppler parameters based on PSO for BFSAR ship target imaging[J]. IEEE Geoscience and Remote Sensing Letters, 2023, 20, 4002605.
4	LI Y M, LIU J, LI X Y, et al. A lightweight network for ship detection in SAR images based on edge feature aware and fusion[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2025, 18, 3782- 3796. doi: 10.1109/JSTARS.2024.3524402
5	FENG Y, ZHANG Y P, ZHANG X Q, et al. Large convolution kernel network with edge self-attention for oriented SAR ship detection[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2025, 18, 2867- 2879. doi: 10.1109/JSTARS.2024.3514855
6	MAN S, YU W D. ELSD-Net: a novel efficient and lightweight ship detection network for SAR images[J]. IEEE Geoscience and Remote Sensing Letters, 2025, 22, 4003505.
7	ZENG T J, ZHANG T W, SHAO Z K, et al. CFAR-DP-FW: a CFAR-guided dual polarization fusion framework for large-scene SAR ship detection[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024, 17, 7242- 7259. doi: 10.1109/JSTARS.2024.3358058
8	吕进东, 王彤, 唐晓斌. 基于图注意力网络的半监督SAR舰船目标检测[J]. 电子与信息学报, 2023, 45 (5): 1541- 1549.
	LV J D, WANG T, TANG X B. Semi-supervised SAR ship target detection with graph attention network[J]. Journal of Electronics & Information Technology, 2023, 45 (5): 1541- 1549.
9	LI C Y, ZHANG Y, SHI Z G, et al. Moderately dense adaptive feature fusion network for infrared small target detection[J]. IEEE Trans. on Geoscience and Remote Sensing, 2024, 62, 5616712.
10	REN S Q, HE K M, GIRSHICK R, et al. Faster R-CNN: 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
11	LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot multibox detector[C]//Proc. of the European Conference on Computer Vision, 2016: 21−37.
12	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: 2999−3007.
13	ZHAO Q, LIU B H, LYU S C, et al. TPH-YOLOv5++: boosting object detection on drone-captured scenarios with cross-layer asymmetric transformer[J]. Remote Sensing, 2023, 15 (6): 1687. doi: 10.3390/rs15061687
14	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.
15	TANG X, ZHANG J Y, XIA Y Z, et al. PEGNet: an enhanced ship detection model for dense scenes and multiscale targets[J]. IEEE Geoscience and Remote Sensing Letters, 2025, 22, 4004405.
16	GAO G, WANG Y J, CHEN Y H, et al. An oriented ship detection method of remote sensing image with contextual global attention mechanism and lightweight task-specific context decoupling[J]. IEEE Trans. on Geoscience and Remote Sensing, 2025, 63, 4200918.
17	SUN Z Z, LENG X G, ZHANG X H, et al. Arbitrary-direction SAR ship detection method for multiscale imbalance[J]. IEEE Trans. on Geoscience and Remote Sensing, 2025, 63, 5208921.
18	CHEN C X, ZHANG Y M, HU R L, et al. A lightweight SAR ship detector using end-to-end image preprocessing network and channel feature guided spatial pyramid pooling[J]. IEEE Geoscience and Remote Sensing Letters, 2024, 21, 4003605.
19	HU B Y, MIAO H X. A lightweight SAR ship detection network based on deep multiscale grouped convolution, network pruning, and knowledge distillation[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2025, 18, 2190- 2207. doi: 10.1109/JSTARS.2024.3502172
20	李毅, 杜兰, 杜宇昂. 基于特征分解卷积神经网络的SAR图像目标检测方法[J]. 雷达学报, 2023, 12 (5): 1069- 1080.
	LI Y, DU L, DU Y A. Convolutional neural network based on feature decomposition for target detection in SAR images[J]. Journal of Radars, 2023, 12 (5): 1069- 1080.
21	TIAN C Y, LIU D C, XUE F L, et al. Faster and lighter: a novel ship detector for SAR images[J]. IEEE Geoscience and Remote Sensing Letters, 2024, 21, 4002005.
22	MIAO T, ZENG H C, YANG W, et al. An improved lightweight RetinaNet for ship detection in SAR images[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2022, 15, 4667- 4679. doi: 10.1109/JSTARS.2022.3180159
29	GHIASI G, CUI Y, SRINIVAS A, et al. Simple copy-paste is a strong data augmentation method for instance segmentation[C]//Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2021: 2917−2927.
30	GUAN Y N, HU H P, LIU W, et al. A false-alarm-controllable modified AdaBoost wake detection method using SAR images[J]. IEEE Sensors Journal, 2023, 23 (23): 29394- 29405. doi: 10.1109/JSEN.2023.3327297
31	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.
32	LI Y H, YAO T, PAN Y W, et al. Contextual transformer networks for visual recognition[J]. IEEE Trans. on Pattern Analysis and Machine Intelligence, 2023, 45 (2): 1489- 1500. doi: 10.1109/TPAMI.2022.3164083
33	VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[EB/OL]. [2025-02-18]. https://arxiv.org/pdf/1706.03762.
34	ZHANG T W, ZHANG X L, LI J W, et al. SAR ship detection dataset （SSDD）: official release and comprehensive data analysis[J]. Remote Sensing, 2021, 13 (18): 3690. doi: 10.3390/rs13183690
35	WANG Y Y, WANG C, ZHANG H, et al. A SAR dataset of ship detection for deep learning under complex backgrounds[J]. Remote Sensing, 2019, 11 (7): 765. doi: 10.3390/rs11070765
36	LIN T Y, DOLLAR 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.
23	WANG F, CHEN C C, ZENG W M. Hybrid multiscale SAR ship detector with CNN-transformer and adaptive fusion loss[J]. IEEE Geoscience and Remote Sensing Letters, 2024, 21, 4015005.
24	ZHOU Y, JIANG X, XU G Z, et al. PVT-SAR: an arbitrarily oriented SAR ship detector with pyramid vision transformer[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2023, 16, 291- 305. doi: 10.1109/JSTARS.2022.3221784
25	严春满, 王铖. 基于选择性坐标注意力的SAR图像舰船目标检测[J]. 电子学报, 2023, 51 (9): 2481- 2491. doi: 10.12263/DZXB.20211416
	YAN C M, WANG C. Ship target detection in SAR image based on selective coordinate attention[J]. Acta Electronica Sinica, 2023, 51 (9): 2481- 2491. doi: 10.12263/DZXB.20211416
26	WU H Y, YU L, LI X W, et al. CTF-Net: a convolutional and transformer fusion network for SAR ship detection[J]. IEEE Geoscience and Remote Sensing Letters, 2023, 20, 4010005.
27	ZHOU S C, ZHANG M, WU L, et al. Lightweight SAR ship detection network based on transformer and feature enhancement[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024, 17, 4845- 4858. doi: 10.1109/JSTARS.2024.3362954
28	CHEN B J, YU C R, ZHAO S, et al. An anchor-free method based on transformers and adaptive features for arbitrarily oriented ship detection in SAR images[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024, 17, 2012- 2028. doi: 10.1109/JSTARS.2023.3325573

层名称	输出大小	网络结构
层名称	输出大小	ResNet-50	ResNet-101	CFN
Conv1	112×112	7×7，64，步长2	7×7，64，步长2	7×7，64，步长2
Conv2_x	56×56	3×3，最大池化，步长2	3×3，最大池化，步长2	3×3，最大池化，步长2
Conv2_x	56×56	$\left[ {\begin{array}{*{20}{c}} {1 \times 1,\;64} \\ {3 \times 3,\;64} \\ {1 \times 1,\;256} \end{array}} \right] \times 3$	$\left[ {\begin{array}{*{20}{c}} {1 \times 1,\;64} \\ {3 \times 3,\;64} \\ {1 \times 1,\;256} \end{array}} \right] \times 3$	$\left[ {\begin{array}{*{20}{c}} {1 \times 1,\;64} \\ {3 \times 3,\;64} \\ {1 \times 1,\;256} \end{array}} \right] \times 3$
Conv3_x	28×28	$ \left[ {\begin{array}{*{20}{c}} {1 \times 1,\;128} \\ {3 \times 3,\;128} \\ {1 \times 1,\;512} \end{array}} \right] \times 4 $	$ \left[ {\begin{array}{*{20}{c}} {1 \times 1,\;128} \\ {3 \times 3,\;128} \\ {1 \times 1,\;512} \end{array}} \right] \times 4 $	$ \left[ {\begin{array}{*{20}{c}} {1 \times 1,\;128} \\ {3 \times 3,\;128} \\ {1 \times 1,\;512} \end{array}} \right] \times 4 $
Conv4_x	14×14	$\left[ {\begin{array}{*{20}{c}} {1 \times 1,\;256} \\ {3 \times 3,\;256} \\ {1 \times 1,\;1024} \end{array}} \right] \times 6$	$\left[ {\begin{array}{*{20}{c}} {1 \times 1,\;256} \\ {3 \times 3,\;256} \\ {1 \times 1,\;1024} \end{array}} \right] \times 23$	$\left[ {\begin{array}{*{20}{c}} {1 \times 1,\;256} \\ {{\text{CoT}},\;256} \\ {1 \times 1,\;1024} \end{array}} \right] \times 6$
Conv5_x	7×7	$\left[ {\begin{array}{*{20}{c}} {1 \times 1,\;512} \\ {3 \times 3,\;512} \\ {1 \times 1,\;2048} \end{array}} \right] \times 3$	$\left[ {\begin{array}{*{20}{c}} {1 \times 1,\;512} \\ {3 \times 3,\;512} \\ {1 \times 1,\;2048} \end{array}} \right] \times 3$	$\left[ {\begin{array}{*{20}{c}} {1 \times 1,\;64} \\ {3 \times 3,\;64} \\ {1 \times 1,\;256} \end{array}} \right] \times 3$

算法	P	R	F₁	AP_0.5	AP_0.75	AP_0.5∶0.95
Faster R-CNN	93.1	90.7	91.8	93.7	77.1	63.2
SSD	88.3	82.8	85.4	88.2	65.1	55.3
YOLOv7	85.8	79.5	82.5	85.5	59.2	52.7
RetinaNet	97.2	83.7	90.0	93.0	76.3	65.0
FFLE-Net	94.9	95.8	95.4	98.5	84.4	71.4

算法	CFN	PAFPN	TCMA	AP_0.5/%	AP_0.75/%	AP_0.5∶0.95/%
消融算法1	−	√	√	95.8	79.1	65.3
FFLE-Net	√	√	√	98.5	84.4	71.4

算法	CFN	PAFPN	TCMA	AP_0.5/%	AP_0.75/%	AP_0.5∶0.95/%
消融算法2	√	−	√	97.4	82.8	69.8
FFLE-Net	√	√	√	98.5	84.4	71.4

算法	CFN	PAFPN	TCMA	AP_0.5/%	AP_0.75/%	AP_0.5∶0.95/%
消融算法3	√	√	−	95.7	75.6	64.4
FFLE-Net	√	√	√	98.5	84.4	71.4