基于渐近式特征融合的轻量化SAR舰船检测算法

doi:10.6040/j.issn.1672-3961.0.2025.099

摘要/Abstract

摘要： 针对合成孔径雷达(synthetic aperture radar, SAR)舰船检测中现有模型感受野受限、多尺度特征融合低效及计算复杂度高的问题,提出通道聚合特征金字塔网络(channel-aggregated feature pyramid network, CA-FPN)。在特征金字塔架构层面,构建跨层级动态加权特征融合机制,通过可学习通道注意力权重自适应校准多分辨率特征,克服传统渐近式特征金字塔(asymptotic feature pyramid network, AFPN)固定采样约束,显著提升多尺度目标表征能力;在特征提取单元设计上,提出并行多尺度空洞卷积模块(parallel multi-scale dilated convolution block, PDBlock)、集成通道注意力(squeeze-and-excitation, SE)与空洞卷积技术,通过特征通道聚合门控机制有效缓解多尺度特征融合过程中的语义冲突。SSDD和LS-SSDD标准数据集上试验评估表明:相较基准模型AFPN, CA-FPN在保持检测精度前提下,模型参数量由1.93×10⁶降至1.17×10⁶(降幅39%),计算复杂度G_FLOPs从4.24降至3.19(降幅24%),SSDD和LS-SSDD数据集上平均精度分别提升2.8%和3.5%,CA-FPN更适应SAR舰船目标检测任务。

关键词: 合成孔径雷达, 舰船检测, 多尺度, 跨层级特征融合, 注意力机制, 空洞卷积

Abstract: To address the limitations of restricted receptive fields, inefficient multi-scale feature fusion, and high computational complexity in existing synthetic aperture radar(SAR)ship detection models, a channel-aggregated feature pyramid network(CA-FPN)was proposed. At the feature pyramid architecture level, a cross-level dynamically weighted feature fusion mechanism was introduced. This mechanism adaptively calibrated multi-resolution features using learnable channel attention weights, which overcame the fixed sampling constraints of the traditional asymptotic feature pyramid network(AFPN)and significantly enhanced multi-scale target representation. In designing the feature extraction unit, a parallel multi-scale dilated convolution block(PDBlock)was developed. By integrating the squeeze-and-excitation(SE)channel attention mechanism with dilated convolution techniques, and employing a feature channel aggregation gating mechanism, the PDBlock effectively mitigated semantic conflicts during multi-scale feature fusion. Experimental evaluations on the standard SSDD and LS-SSDD datasets demonstrated that, compared to the baseline AFPN model, CA-FPN maintained detection accuracy while reducing model parameters from 1.93×10⁶ to 1.17×10⁶(a 39% reduction)and computational complexity(in G_FLOPs)from 4.24 to 3.19(a 24% reduction). The mean average precision increased by 2.8% on SSDD and 3.5% on LS-SSDD. CA-FPN was more effective and better adapted to the requirements of SAR ship target detection tasks.

Key words: synthetic aperture radar, ship detection, multi-scale, cross-level feature fusion, attention mechanism, dilated convolution

中图分类号:

TP183

刘飞宇,张静,王亦楠. 基于渐近式特征融合的轻量化SAR舰船检测算法[J]. 山东大学学报 (工学版), 2026, 56(2): 52-59.

LIU Feiyu, ZHANG Jing, WANG Yinan. Lightweight SAR ship detection algorithm based on asymptotic feature fusion[J]. Journal of Shandong University(Engineering Science), 2026, 56(2): 52-59.

参考文献

[1] DONG G G, KUANG G Y, WANG N, et al. Classification via sparse representation of steerable wavelet frames on Grassmann manifold: application to target recognition in SAR image[J]. IEEE Transactions on Image Processing, 2017, 26(6): 2892-2904.
[2] 张弘森, 徐建, 吴飞, 等. 基于深度学习的视觉图像舰船目标检测与识别方法综述[J]. 指挥信息系统与技术, 2024, 15(6):48-63. ZHANG Hongsen, XU Jian, WU Fei, et al. Review of ship target detection and recognition methods with visual imagergy based on deep learning[J].Command Information System and Technology, 2024, 15(6): 48-63.
[3] 张尧, 张艳, 王涛, 等. 大场景SAR影像舰船目标检测的轻量化研究[J]. 地球信息科学学报, 2025, 27(1): 256-270. ZHANG Yao, ZHANG Yan, WANG Tao, et al. Lightweight research on ship target detection in large-scale SAR images[J]. Journal of Geo-Information Science, 2025, 27(1): 256-270.
[4] 张潋钟. SAR图像舰船目标快速检测识别技术[D]. 成都:电子科技大学, 2022: 2-5. ZHANG Lianzhong. Rapid detection and recognition technology of ship targets in SAR images[D]. Chengdu: University of Electronic Science and Technology of China, 2022: 2-5.
[5] TANG X, ZHANG J F, XIA Y Z, et al. DBW-YOLO: a high-precision SAR ship detection method for complex environments[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024, 17: 7029-7039.
[6] CHEN S Q, ZHAN R H, WANG W, et al. Learning slimming SAR ship object detector through network pruning and knowledge distillation[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2021, 14: 1267-1282.
[7] WANG Y T, MIAO M, ZHU S L. L-ENet: an ultra-lightweight SAR image detection network[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024, 17: 8967-8978.
[8] TAN X D, LENG X G, SUN Z Z, et al. Lightweight ship detection network for SAR range-compressed domain[J]. Remote Sensing, 2024, 16(17): 3284.
[9] YU H, YANG S H, ZHOU S P, et al. VS-LSDet: a multiscale ship detector for spaceborne SAR images based on visual saliency and lightweight CNN[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024, 17: 1137-1154.
[10] 陈志文, 王坤, 周广蕴, 等. 基于胶囊网络及其权重剪枝的SAR图像变化检测方法[J]. 计算机科学, 2021, 48(7): 190-198. CHEN Zhiwen, WANG Kun, ZHOU Guangyun, et al. SAR image change detection method based on capsule network with weight pruning[J]. Computer Science, 2021, 48(7): 190-198.
[11] YANG G Y, LEI J, ZHU Z K, et al. AFPN: asymptotic feature pyramid network for object detection[C] //2023 IEEE International Conference on Systems, Man, and Cybernetics(SMC). Honolulu, USA: IEEE, 2023: 2184-2189.
[12] HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C] //2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, USA: IEEE, 2016: 770-778.
[13] LI S Y, WANG Z D, LIU Z C, et al. MogaNet: multi-order gated aggregation network[EB/OL].(2025-06-16)[2025-06-25]. https://arxiv.org/abs/2211.03295
[14] HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C] //2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, USA: IEEE, 2018: 7132-7141.
[15] QIAO Y L, GUO Y Y, HE D J. Cattle body detection based on YOLOv5-ASFF for precision livestock farming[J]. Computers and Electronics in Agriculture, 2023, 204: 107579.
[16] LIU Z, LIN Y T, CAO Y, et al. Swin transformer: hierarchical vision transformer using shifted windows[C] //2021 IEEE/CVF International Conference on Computer Vision. Montreal, Canada: IEEE, 2021: 10012-10022.
[17] LUO W, LI Y, URTASUN R, et al. Understanding the effective receptive field in deep convolutional neural networks[C] //Proceedings of the 31st International Conference on Neural Information Processing Systems. Long Beach, USA: Curran Associates Inc., 2017: 4905-4914.
[18] VASWANI A, RAMACHANDRAN P, SRINIVAS A, et al. Scaling local self-attention for parameter efficient visual backbones[C] //2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville, USA: IEEE, 2021: 12889-12899.
[19] ZHANG T W, ZHANG X L, KE X, et al. LS-SSDD-v1.0: a deep learning dataset dedicated to small ship detection from large-scale sentinel-1 SAR images[J]. Remote Sensing, 2020, 12(18): 2997.
[20] 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.
[21] YU G H, CHANG Q Y, LV W Y, et al. PP-PicoDet: a better real-time object detector on mobile devices[EB/OL].(2021-11-01)[2025-06-25]. https://arxiv.org/abs/2111.00902
[22] WANG A, CHEN H, LIU L, et al. Yolov10: real-time end-to-end objectdetection[EB/OL].(2024-10-30)[2025-06-25]. https://arxiv.org/abs/2405.14458
[23] KHANAM R, HUSSAIN M. YOlOv11: an overview of the key architectural enhancements[EB/OL].(2024-10-23)[2025-06-25]. https://www.arxiv.org/abs/2410.17725
[24] HOWARD A, SANDLER M, CHEN B, et al. Searching for MobileNetV3[C] //2019 IEEE/CVF International Conference on Computer Vision. Seoul: IEEE, 2019: 1314-1324.
[25] FENG Y, CHEN J, HUANG Z X, et al. A lightweight position-enhanced anchor-free algorithm for SAR ship detection[J]. Remote Sensing, 2022, 14(8): 1908.
[26] YU N J, REN H H, DENG T M, et al. A lightweight radar ship detection framework with hybrid attentions[J]. Remote Sensing, 2023, 15(11): 2743.
[27] 唐志勇, 魏雪云, 江蒋伟, 等. 基于YOLO框架的轻量化SAR图像舰船检测方法研究[J]. 无线电工程, 2024, 54(10): 2347-2354. TANG Zhiyong, WEI Xueyun, JIANG Jiangwei, et al. Research on lightweight SAR image ship detection method based on YOLO framework[J].Radio Engineering, 2024, 54(10): 2347-2354.
[28] SUN Z Z, LENG X G, ZHANG X H, et al. Arbitrary-direction SAR ship detection method for multiscale imbalance[J]. IEEE Transactions on Geoscience and Remote Sensing, 2025, 63: 5208921.
[29] CAO J, HAN P H, LIANG H P, et al. SFRT-DETR:a SAR ship detection algorithm based on feature selection and multi-scale feature focus[J]. Signal, Image and Video Processing, 2024, 19(2): 115.
[30] TAN M X, PANG R M, LE Q V. Efficientdet: scalable and efficient object detection[C] //2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle, USA: IEEE, 2020: 10778-10787.
[31] TANG H D, GAO S, LI S, et al. A lightweight SAR image ship detection method based on improved convolution and YOLOv7[J]. Remote Sensing, 2024, 16(3): 486.
[32] LIU Y B, YAN G, MA F, et al. SAR ship detection based on explainable evidence learning under intraclass imbalance[J]. IEEE Transactions on Geoscience and Remote Sensing, 2024, 62: 5207715.
[33] XU X W, ZHANG X L, ZHANG T W, et al. On-board ship detection in SAR images based on L-YOLO[C] //2022 IEEE Radar Conference(RadarConf22). New York, USA: IEEE, 2022: 1-5.

多维度评价

Viewed

Full text

Abstract

Cited

Shared

Discussed