基于深度学习的SAR图像海洋涡旋自动检测模型

doi:10.11978/2024242

热带海洋学报 ›› 2025, Vol. 44 ›› Issue (5): 65-76.doi: 10.11978/2024242CSTR: 32234.14.2024242

基于深度学习的SAR图像海洋涡旋自动检测模型

刘太龙¹(), 谢涛¹^,²^,³^,⁴(), 李建¹^,³^,⁴, 王超¹^,⁵, 张雪红¹^,³^,⁴

1.南京信息工程大学遥感与测绘工程学院, 江苏南京 210044
2.青岛海洋科技中心区域海洋动力学与数值模拟功能实验室, 青岛山东 266200
3.自然资源部遥感导航一体化应用工程技术创新中心, 江苏南京 210044
4.江苏省协同精密导航定位与智能应用工程研究中心, 江苏南京 210044
5.南京信息工程大学电子与信息工程学院, 江苏南京 210044

收稿日期:2024-12-29 修回日期:2025-02-18 出版日期:2025-09-10 发布日期:2025-10-14
通讯作者: 谢涛
作者简介:
刘太龙(1998—), 男, 硕士研究生, 主要从事中尺度涡方面的研究。email: 13696350617@163.com
基金资助:
国家重点研发计划项目(2022YFC3104900); 国家重点研发计划项目(2022YFC3104905); 国家自然科学基金项目(42176180)

Deep learning-based automatic detection model for ocean eddies in SAR images

LIU Tailong¹(), XIE Tao¹^,²^,³^,⁴(), LI Jian¹^,³^,⁴, WANG Chao¹^,⁵, ZHANG Xuehong¹^,³^,⁴

1. School of Remote Sensing & Geomatics Engineering, Nanjing University of Information Science & Technology, Nanjing 210044, China
2. Laboratory for Regional Oceanography and Numerical Modeling, Qingdao Marine Science and Technology Center, Qingdao 266200, China
3. Innovation Center for Integrated Remote Sensing and Navigation Applications Engineering Technology, Ministry of Natural Resources, Nanjing 210044, China
4. Jiangsu Collaborative Innovation Center for Precision Navigation and Intelligent Application Engineering, Nanjing 210044, China
5. School of Electronic and Information Engineering, Nanjing University of Information Science and Technology, Nanjing 210044, China

Received:2024-12-29 Revised:2025-02-18 Online:2025-09-10 Published:2025-10-14
Contact: XIE Tao
Supported by:
National Key Research and Development Program of China(2022YFC3104900); National Key Research and Development Program of China(2022YFC3104905); National Natural Science Foundation of China(42176180)

摘要/Abstract

摘要：

传统基于合成孔径雷达(synthetic aperture radar, SAR)数据的涡旋检测方法需手动设定阈值和特征参数, 操作复杂且难以自动化, 现有深度学习模型在检测过程中也存在较高的漏检和误检情况, 难以满足涡旋检测对精度和效率的要求。为解决这些问题, 文章提出一种基于YOLOv8 (you only look once version 8)的改进模型, 即涡旋检测网络(eddy detection network, EddyDetNet), 以克服上述局限性, 提升检测的准确性和效率。该模型在骨干网络(Backbone)和颈部网络(Neck)中引入自适应特征融合模块(adaptive feature compression module, AFCM)和多尺度特征空间金字塔模块(multi-scale feature spatial pyramid module, MFSP)并优化Neck结构和在头部网络(Head)部分新增小目标检测头, 提升了不同尺度涡旋的检测精度。实验结果表明, EddyDetNet在精确率(precision, p)、召回率(recall, r)和平均精确率均值(mean average precision, mAP)指标上较YOLOv8分别提升了2.4%、3.2%和5.5%, 参数量和运算量减少了38.1%和15.8%。与YOLOv8相比, EddyDetNet在降低计算复杂度和参数量的同时, 保持了较高检测精度, 适用于多目标和复杂背景下的涡旋检测任务。

关键词: 海洋涡旋, 合成孔径雷达, 深度学习, YOLO, 目标检测

Abstract:

Traditional eddy detection methods based on SAR data require manual threshold setting and feature parameter tuning, making processes complex and difficult to automate. Existing deep learning models also suffer from high rates of false negatives and false positives, failing to meet the accuracy and efficiency requirements of eddy detection. To address these issues, this paper proposes an improved model based on YOLOv8 (you only look once version 8), named EddyDetNet, to overcome the above limitations and enhance both detection accuracy and efficiency. This model introduces an adaptive feature compression module (AFCM) and a multi-scale feature spatial pyramid module (MFSP) in Backbone and Neck, optimizes the Neck structure, and adds a small target detection head in the Head part, thereby improving the detection accuracy of vortices of different scales. Experimental results show that EddyDetNet outperforms YOLOv8 by 2.4%, 3.2%, and 5.5% in precision (p), recall (r), and mean Average Precision (mAP), respectively, while reducing parameter size and computational complexity by 38.1% and 15.8%. Compared to YOLOv8, EddyDetNet reduces computational complexity and parameter size while maintaining high detection accuracy, making it suitable for eddy detection tasks in multi-target and complex background scenarios.

Key words: oceanic eddies, synthetic aperture radar, deep learning, YOLO, object detection

中图分类号:

P731.27

刘太龙, 谢涛, 李建, 王超, 张雪红. 基于深度学习的SAR图像海洋涡旋自动检测模型[J]. 热带海洋学报, 2025, 44(5): 65-76.

LIU Tailong, XIE Tao, LI Jian, WANG Chao, ZHANG Xuehong. Deep learning-based automatic detection model for ocean eddies in SAR images[J]. Journal of Tropical Oceanography, 2025, 44(5): 65-76.

图/表 9

图1

图2

图3

图4

图5

表1

表2

图6

图7

参考文献 29

[1]	陈戈, 杨杰, 田丰林, 等, 2021. 海洋涡旋遥感: 进展与挑战[J]. 遥感学报, 25(1): 302-322.
	CHEN GE, YANG JIE, TIAN FENGLIN, et al, 2021. Remote sensing of oceanic eddies: progresses and challenges[J]. National Remote Sensing Bulletin, 25(1): 302-322 (in Chinese with English abstract).
[2]	董昌明, 2015. 海洋涡旋探测与分析[M]. 北京: 科学出版社.
	DONG CHANGMING, 2015. Oceanic eddy detection and analysis[M]. Beijing: Science Press (in Chinese).
[3]	甘滢晖, 2022. 基于遥感资料的中尺度涡与叶绿素分布的关系研究[D]. 大连: 大连海洋大学.
	GAN YINGHUI, 2022. Study on the relationship between mesoscale Vortex and chlorophyll distribution based on remote sensing data[D]. Dalian: Dalian Ocean University (in Chinese with English abstract).
[4]	高山, 王凡, 李明悝, 等, 2007. 中尺度涡的高度计资料同化模拟[J]. 中国科学(D辑: 地球科学), 37(12): 1669-1678.
	GAO SHAN, WANG FAN, LI MINGKUI, et al, 2007. Assimilation simulation of mesoscale Vortex with altimeter data[J]. Science in China (Series D: Earth Sciences), 37(12): 1669-1678 (in Chinese).
[5]	沈飙, 陈扬, 杨琛, 等, 2020. 海洋科学中尺度涡的计算机视觉检测和分析方法[J]. 数据与计算发展前沿, 2(6): 30-41.
	SHEN BIAO, CHEN YANG, YANG CHEN, et al, 2020. Computer vision detection and analysis of mesoscale eddies in marine science[J]. Frontiers of Data & Computing, 2(6): 30-41 (in Chinese with English abstract).
[6]	吴进群, 陈戈, 马纯永, 等, 2023. 深度学习的SAR图像海洋涡旋自动检测及其特征提取[J]. 舰船科学技术, 45(6): 170-173.
	WU JINQUN, CHEN GE, MA CHUNYONG, et al, 2023. Research on automatic detection of oceanic eddies in SAR image and extraction of its characteristic parameters based on depth learning[J]. Ship Science and Technology, 45(6): 170-173 (in Chinese with English abstract).
[7]	赵文涛, 俞建成, 张艾群, 等, 2016. 基于卫星测高数据的海洋中尺度涡流动态特征检测[J]. 海洋学研究, 34(3): 62-68. doi: 10.3969/j.issn.1001-909X.2016.03.010
	ZHAO WENTAO, YU JIANCHENG, ZHANG AIQUN, et al, 2016. Dynamic feature detection of mesoscale eddies based on SLA data[J]. Journal of Marine Sciences, 34(3): 62-68 (in Chinese with English abstract).
[8]	郑晓莉, 董庆, 樊星, 2020. 北太平洋中尺度涡海表温度和叶绿素浓度特征分析[J]. 遥感学报, 24(1): 85-96.
	ZHENG XIAOLI, DONG QING, FAN XING, 2020. Characteristics of sea surface temperature and Chlorophyll concentration inside mesoscale eddies in the North Pacific Ocean[J]. Journal of Remote Sensing, 24(1): 85-96 (in Chinese with English abstract).
[9]	ALPERS W, BRANDT P, LAZAR A, et al, 2013. A small-scale oceanic eddy off the coast of West Africa studied by multi-sensor satellite and surface drifter data[J]. Remote Sensing of Environment, 129: 132-143.
[10]	ALPERS W R, ROSS D B, RUFENACH C L, 1981. On the detectability of ocean surface waves by real and synthetic aperture radar[J]. Journal of Geophysical Research: Oceans, 86(C7): 6481-6498.
[11]	CAZENAVE A, GOUZENES Y, BIROL F, et al, 2022. Sea level along the world’s coastlines can be measured by a network of virtual altimetry stations[J]. Communications Earth & Environment, 3: 117.
[12]	CHELTON D B, SCHLAX M G, SAMELSON R M, et al, 2007. Global observations of large oceanic eddies[J]. Geophysical Research Letters, 34(15): L15606.
[13]	CHEN JIERUN, KAO S H, HE HAO, et al, 2023. Run, don’t walk: chasing higher FLOPS for faster neural networks[C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Vancouver, BC: IEEE: 12021-12031.
[14]	DONG CHANGMING, MCWILLIAMS J C, LIU YU, et al, 2014. Global heat and salt transports by eddy movement[J]. Nature Communications, 5: 3294. doi: 10.1038/ncomms4294 pmid: 24534770
[15]	HUANG DONGMEI, DU YANLING, HE QI, et al, 2017. DeepEddy: a simple deep architecture for mesoscale oceanic eddy detection in SAR images[C]//2017 IEEE 14th International Conference on Networking, Sensing and Control (ICNSC). Calabria: IEEE: 673-678.
[16]	JOCHER G, QIU J, CHAURASIA A, 2023. Ultralytics YOLO (Version 8.0.0)[Computer software]. https://github.com/ultralytics.
[17]	KHACHATRIAN E, SANDALYUK N, LOZOU P, 2023. Eddy detection in the marginal ice zone with sentinel-1 data using YOLOv5[J]. Remote Sensing, 15(9): 2244.
[18]	LIU WEI, ANGUELOV D, ERHAN D, et al, 2016. SSD: single shot MultiBox detector[C]// Computer Vision - ECCV 2016. Cham: Springer International Publishing: 21-37.
[19]	OUYANG DALIANG, HE SU, ZHANG GUOZHONG, et al, 2023. Efficient multi-scale attention module with cross-spatial learning[C]// ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). Rhodes Island: IEEE: 1-5.
[20]	REDMON J, DIVVALA S, GIRSHICK R, et al, 2016. You only look once: unified, real-time object detection[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas, NV: IEEE: 779-788.
[21]	REN SHAOQING, HE KAIMING, GIRSHICK R, et al, 2017. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 39(6): 1137-1149. doi: 10.1109/TPAMI.2016.2577031 pmid: 27295650
[22]	RIZAEV I G, KARAKUŞ O, HOGAN S J, et al, 2022. Modeling and SAR imaging of the sea surface: a review of the state-of-the-art with simulations[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 187: 120-140.
[23]	WAN DAHANG, LYU RONGSHENG, SHEN SIYUAN, et al, 2023. Mixed local channel attention for object detection[J]. Engineering Applications of Artificial Intelligence, 123: 106442.
[24]	WANG C Y, BOCHKOVSKIY A, LIAO H M, 2023. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Vancouver, BC: IEEE: 7464-7475.
[25]	WANG C Y, YEH I H, MARK LIAO H Y, 2024. Yolov9: Learning what you want to learn using programmable gradient information[C]// European conference on computer vision. Cham: Springer Nature Switzerland: 1-21.
[26]	XIA LINGHUI, CHEN GE, CHEN XIAOYAN, et al, 2022. Submesoscale oceanic eddy detection in SAR images using context and edge association network[J]. Frontiers in Marine Science, 9: 1023624.
[27]	XU MING, LI HONGPING, YUN YUYING, et al, 2023. End-to-end pixel-wisely detection of oceanic eddy on SAR images with stacked attention network[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 16: 9711-9724.
[28]	ZHANG DI, GADE M, WANG WENSHENG, et al, 2023. EddyDet: a deep framework for oceanic eddy detection in synthetic aperture radar images[J]. Remote Sensing, 15(19): 4752.
[29]	ZI NANNAN, LI XIAOMING, GADE M, et al, 2024. Ocean eddy detection based on YOLO deep learning algorithm by synthetic aperture radar data[J]. Remote Sensing of Environment, 307: 114139.

改进方法	AFCM	MFSP	Neck+Head	p/%	r/%	mAP@0.5/%	参数量/M	GFLOPs
1				88.2	88.6	89.5	24.64	78.7
2	√			86.2	92.5	94.7	15.85	50.2
3		√		88.0	89.2	92.5	24.84	80.2
4			√	88.9	91.0	93.4	24.16	99.6
5	√	√	√	90.6	91.8	95.0	15.26	66.3

模型名称	AP/%(AE)	AP/%(CE)	p/%	r/%	mAP@0.5/%
SSD	47.01	81.24	58.48	56.90	64.12
Faster R-CNN	32.99	79.01	43.67	73.76	56.00
YOLOv7	42.60	66.30	51.90	61.10	54.50
YOLOv8	85.20	93.70	88.20	88.60	89.50
YOLOv9	39.70	78.80	52.30	76.00	59.20
EddyDetNet	94.20	95.80	90.60	91.80	95.00

基于深度学习的SAR图像海洋涡旋自动检测模型

Deep learning-based automatic detection model for ocean eddies in SAR images

RichHTML

PDF (PC)

赞

可视化

摘要/Abstract

引用本文

使用本文

图/表 9

参考文献 29

相关文章 5

Metrics

本文评价

推荐阅读 0

[1]	李天阔, 屈科, 李晓涵, 王傲宇, 王超. 半潜式浮式风机基础水动力特性及系泊系统张力特性预测[J]. 热带海洋学报, 2025, 44(3): 36-47.
[2]	高榕泽, 屈科, 任兴月, 王旭. 卷积神经网络方法在岛礁类海啸波水动力特性演变的应用[J]. 热带海洋学报, 2024, 43(4): 68-75.
[3]	赵杰, 林延奖, 刘燃, 杜榕. 基于PredRNN++模型对南海中尺度涡旋的预测研究[J]. 热带海洋学报, 2024, 43(1): 16-27.
[4]	刘爽, 经志友, 詹海刚. 基于生成对抗网络模型的热带和亚热带海洋中尺度涡预报研究[J]. 热带海洋学报, 2022, 41(5): 1-16.
[5]	巩彪, 黄韦艮, 陈鹏. 基于改进的Hough变换ASAR图像船只尾迹自动识别算法[J]. 热带海洋学报, 2014, 33(3): 33-39.