A Cross-modal Style Diffusion Sonar Image Synthesis Method Based on Physical Gidance and Structural Constraints

ZHAO Qiaoqiao; ZHANG Lichuan; LI Xuanfeng; PAN Guang

doi:10.12382/bgxb.2025.1133

您当前的位置：

首页 >

文章列表页 >

A Cross-modal Style Diffusion Sonar Image Synthesis Method Based on Physical Gidance and Structural Constraints

更新时间：2026-08-25

- A Cross-modal Style Diffusion Sonar Image Synthesis Method Based on Physical Gidance and Structural Constraints
- Acta Armamentarii (2026)
- 作者机构：
  
  西北工业大学航海学院, 陕西西安 710072
- 作者简介：
- 基金信息：
- DOI：10.12382/bgxb.2025.1133
  CLC： TP535
- Received：23 December 2025，
  
  Revised：2026-08-24，
  
  Accepted：09 July 2026，
- 稿件说明：
移动端阅览
ZHAO Qiaoqiao, ZHANG Lichuan, LI Xuanfeng, et al. A Cross-modal Style Diffusion Sonar Image Synthesis Method Based on Physical Gidance and Structural Constraints[J/OL]. Acta Armamentarii, 2026.
DOI：

ZHAO Qiaoqiao, ZHANG Lichuan, LI Xuanfeng, et al. A Cross-modal Style Diffusion Sonar Image Synthesis Method Based on Physical Gidance and Structural Constraints[J/OL]. Acta Armamentarii, 2026. DOI： 10.12382/bgxb.2025.1133.

摘要

为解决水下声呐感知中高质量标注数据稀缺的难题，建立一种基于物理引导和结构约束的扩散模型声呐图像合成方法，用于光学遥感到声呐图像的跨模态合成

。现有扩散模型方法多针对通用光学图像，在处理遥感光学至声呐图像跨模态任务时，往往忽视声呐成像机理，导致声学特性退化及几何结构崩溃。为此，构建频域噪声对齐和声学先验注入模块，在扩散去噪的初始阶段即嵌入真实声呐纹理分布与高光阴影物理先验；同时，引入结构约束的注意力机制，对目标和背景区域实施差异化风格迁移策略，在生成逼真声学特征的同时保留目标的几何轮廓与位置结构，实现光学遥感标签的直接迁移。实验结果表明，该方法在声学真实性和目标结构性上更优。利用生成的数据辅助训练YOLOv8检测器，0.50~0.95区间内的平均精度均值

mAP

50-95

从0.724提升至0.771。所建立方法为缓解水下声呐感知数据匮乏问题提供了一种实用的工程化解决方案。

Abstract

To address the challenge of scarce high-quality annotated data in underwater sonar perception

a cross-modal style diffusion sonar image synthesis method based on physical guidance and structural constraints is constructed for the synthesis of sonar images from optical remote sensing. The existing diffusion model methods are mostly designed for generic optical images. When dealing with the cross-modal task of converting the remote sensing optical images to the sonar images

the acoustic imaging mechanism is often overlooked

which leads to the degradation of acoustic characteristics and the collapse of geometric structures. To tackle this issue

a frequency-domain noise alignment module and an acoustic prior injection module are constructed. These modules embed the texture distribution of real sonar images and the physical priors of highlight-shadow relationships at the initial stage of diffusion denoising. Meanwhile

an attention mechanism with structural constraints is introduced

which implements a differentiated style transfer strategy for target and background regions. This method ensures the generation of realistic acoustic features while preserving the geometric contours and positional structures of targets

thus enabling the direct transfer of labels from optical remote sensing images. Experimental results demonstrate that the proposed method is superior in terms of both acoustic realism and structural integrity. The

mAP

50-95

is significantly improved from 0.724 to 0.771 by using the generated data to assist in training the YOLOv8 detector. This study provides a p

ractical engineering solution for alleviating the problem of data scarcity in underwater sonar perception.

关键词

Keywords

references

Qin K S , Liu D , Wang F , et al . Improved YOLOv7 model for underwater sonar image object detection [J ] . Journal of Visual Communication and Image Representation , 2024 , 100 : 104124 .

Chen L , Huang Y Z , Dong J Y , et al . Underwater optical object detection in the era of artificial intelligence: Current, challenge, and future [J ] . ACM Computing Surveys , 2025 , 58 ( 3 ): 1 - 34 .

Aubard M , Madureira A , Teixeira L , et al . Sonar-based deep learning in underwater robotics: overview, robustness, and challenges [J ] . IEEE Journal of Oceanic Engineering , 2025 , 50 ( 3 ): 1866 - 1884 .

Huang C , Zhao J H , Zhang H M , et al . Seg2Sonar: a full-class sample synthesis method applied to underwater sonar image target detection, recognition, and segmentation tasks [J ] . IEEE Transactions on Geoscience and Remote Sensing , 2024 , 62 : 5909319 .

Denos K , Ravaut M , Fagette A , et al . Deep learning applied to underwater mine warfare [C ] // OCEANS Aberdeen , 2017 : 1 - 7 .

Jiang Y F , Ku B W , Kim W J , et al . Side-scan sonar image synthesis based on generative adversarial network for images in multiple frequencies [J ] . IEEE Geoscience and Remote Sensing Letters , 2021 , 18 ( 9 ): 1505 - 1509 .

Zhou J Q , Li Y , Qin H M , et al . Sonar image generation by MFA-CycleGAN for boosting underwater object detection of AUVs [J ] . IEEE Journal of Oceanic Engineering , 2024 , 49 ( 3 ): 905 - 919 .

ho J , jain A N , Abbeel P . Denoising diffusion probabilistic models [J ] . Advances in Neural Information Processing Systems (NeurIPS) , 2020 , 33 : 6840 - 6851 .

Yang Z W , ZHAO J H , ZHANG H M , et al . A side-scan sonar image synthesis method based on a diffusion model [J ] . Journal of Marine Science and Engineering , 2023 , 11 ( 6 ): 1103 .

Natarajan P , Basha K , Nambiar A . Synth-sonar: Sonar image synthesis with enhanced diversity and realism via dual diffusion models and GPT prompting [PP/OL ] . arXiv ( 2024-10-11 ). https://arxiv.org/abs/2410.08612 https://arxiv.org/abs/2410.08612 .

Yang Z W , Zhao J H , Yu Y C , et al . A sample augmentation method for side-scan sonar full-class images that can be used for detection and segmentation [J ] . IEEE Transactions on Geoscience and Remote Sensing , 2024 , 62 : 5908111 .

Wen X , Zhang F H , Cheng C S , et al . Side-scan sonar underwater target detection: Combining the diffusion model with an improved YOLOv7 model [J ] . IEEE Journal of Oceanic Engineering , 2024 , 49 ( 3 ): 976 - 991 .

Chung J , Hyun S H , Heo J H . Style injection in diffusion: a training-free approach for adapting large-scale diffusion models for style transfer [C ] // IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024 : 8795 - 8805 .

张鹏飞 , 刘维 , 江泽林 , 等 . 合成孔径声纳图像阴影增强方法研究 [J ] . 兵工学报 , 2015 , 36 ( 2 ): 305 - 312 .

Zhang P F , Liu W , Jiang Z L , et al . Research on shadow enhancement for synthetic aperture sonar images [J ] . Acta Armamentarii , 2015 , 36 ( 2 ): 305 - 312 . (in Chinese) .

武鹤龙 , 邱政 , 张维全 . 侧扫声纳图像非下采样轮廓波变换域分区增强方法 [J ] . 兵工学报 , 2021 , 42 ( 7 ): 1463 - 1470 .

Wu H L , Qiu Z , Zhang W Q . Partition enhancement method for nsct domain of side-scan sonar image [J ] . Acta Armamentarii , 2021 , 42 ( 7 ): 1463 - 1470 . (in Chinese) .

Rombach R , Blattmann A , Lorenz D , et al . High-resolution image synthesis with latent diffusion models [C ] // IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022 : 10684 - 10695 .

Hu E J , Shen Y L , Wallis P , et al . LoRA: low-rank adaptation of large language models [C ] // International Conference on Learning Representations . 2022 .

Hertz A , Mokady R , Tenenbaum J , et al . Prompt-to-prompt image editing with cross attention control [PP/OL ] . arXiv ( 2022-08-02 ). https://arxiv.org/abs/2208.01626 https://arxiv.org/abs/2208.01626 .

Zhang Z S , Zhang P , Wang F X , et al . Self-supervised enhancement of forward-looking sonar images: Bridging cross-modal degradation gaps through feature space transformation and multi-frame fusion [PP/OL ] . arXiv ( 2025 - 04 - 15 ). https://arxiv.org/abs/2504.10974 https://arxiv.org/abs/2504.10974 .

Liu Z K , Yuan L , Weng L B , et al . A high resolution optical satellite image dataset for ship recognition and some new baselines [C ] // The 6th International Conference on Pattern Recognition Applications and Methods , 2017 , 1 : 324 - 331 .

Cheng G , Han J W , Lu X Q . Remote sensing image scene classification: Benchmark and state of the art [J ] . Proceedings of the IEEE , 2017 , 105 ( 10 ): 1865 - 1883 .

Huo G Y , Wu Z Y , Li J B . Underwater object classification in side-scan sonar images using deep transfer learning and semisynthetic training data [J ] . IEEE Access , 2020 , 8 : 47407 - 47418 .

Wang Z , Bovik A C , Sheikh H R , et al . Image quality assessment: from error visibility to structural similarity [J ] . IEEE Transactions on Image Processing , 2004 , 13 ( 4 ): 600 - 612 .

Zhu J Y , Park T , Isola P , et al . Unpaired image-to-image translation using cycle-consistent adversarial networks [C ] // 2017 IEEE International Conference on Computer Vision , 2017 : 2242 - 2251 .

Views

下载量

CNKI被引量

Alert me when the article has been cited

提交

Tools

Publicity Resources

Machine Learning-based Models for Predicting the Penetration Depth of Concrete

Research on Surface Quality Diagnosis of Magnesium Alloy Deposited Layers in CMT Wire-arc Additive Manufacturing Process Based on SSA-LSTM Network

The Influence of Arc Additive and Subtractive Hybrid Manufacturing on the Forming Quality of 316L Stainless Steel

Research on the Influence of Conical Angle Difference on the Stability of Sealing Performance of Conical Pipe Joints under Vibration Load

Research on Additive Friction Stir Deposition Process of Armored Aluminum Alloy

Related Author

Meng LI

Haijun WU

Heng DONG

Guang REN

Peng ZHANG

Fenglei HUANG

Related Institution

State Key Laboratory of Explosive Science and Technology, Beijing Institute of Technology

No.660 Institute of Hongdu Aviation Industry, AVIC

Postal code：100089
Tel：010-68963060/68962718 Fax：010-68963025 Email：bgxb@cos.org.cn
Technical support is provided by Beijing Founder electronics co., LTD 京ICP备05059581号-4 京公网安备11010802024360号
It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
Total Visits：261136 Visits Today：1331 Online Users：5

⁰