Evaluation Method of Marine Biofouling Based on Deep Learning

ZENG Chaofan; JIAO Jianbo; OU Zhihui; CHEN Baiyi; YANG Yanling; WU Jianhua

doi:10.16490/j.cnki.issn.1001-3660.2025.24.012

PDF(11362 KB)

Surface Technology ›› 2025, Vol. 54 ›› Issue (24) : 149-162. DOI: 10.16490/j.cnki.issn.1001-3660.2025.24.012

Corrosion and Protection

Evaluation Method of Marine Biofouling Based on Deep Learning

ZENG Chaofan^1,2, JIAO Jianbo^1,2, OU Zhihui^1,2, CHEN Baiyi^1,2, YANG Yanling^1,2,*, WU Jianhua^1,2,*

Author information +

History +

Abstract

Marine biofouling poses significant adverse effects on human marine activities. As biofouling adheres to the surface of ship hulls and marine structures, increasing navigation resistance and fuel consumption, incurring substantial economic losses, objectively and accurately assessing the degree of biofouling is a pivotal factor for effective fouling prevention. Traditional evaluation methods, such as manual visual inspections by professionals, suffer from inefficiency and subjectivity. Conventional image processing techniques exhibit low accuracy in complex marine environments. To address these limitations, the work aims to develop a fast, robust and high-precision intelligent method for quantitative evaluation of marine biofouling.
An enhanced YOLO algorithm, YOLO-FPSC was proposed. Combined with network redesign, targeted attention modeling and loss-function, the algorithm was validated through systematic empirical evaluation and ablation studies. The YOLO-FPSC focused on three key aspects. Feature extraction enhancement, to resolve issues of image blur and low contrast, the Spatial Pyramid Pooling-Fast (SPPF) module in the Backbone was replaced with a Feature Pyramid Shared Convolution (FPSC). This modification enhanced the capture of local image details through multi-scale feature fusion while reducing noise interference. The FPSC employed shared convolutional kernels with varying dilation rates to expand the receptive field without increasing computational complexity, ensuring robust feature representation under suboptimal imaging conditions. Attention-Driven Feature Refinement: an Adaptive Fine-Grained Classification (AFGC) attention mechanism was integrated into the Neck layer to dynamically focus on critical regions of interest. By simulating human visual attention through high, medium, and low-resolution feature interactions, the AFGC mechanism amplified the discriminative power of surface textures and edge details of biofouling, effectively minimizing missed detection and false positives. To address boundary ambiguity and overlapping biofouling clusters, the conventional Complete Intersection over Union (CIoU) loss was replaced with the Normalized Wasserstein Distance (NWD) loss. NWD modeled bounding boxes as Gaussian distributions and measured their similarity with Wasserstein distance, providing smoother gradient signals and improved robustness for small and overlapping targets. Experimentally, YOLO-FPSC was trained and evaluated on a dedicated annotated marine biofouling image dataset. Data included stratified train, validation and test splits, targeted data augmentations, and comparisons against baseline models.
The improved algorithm YOLO-FPSC on biofouling dataset achieved mean average precision (mAP) of 96.2%, surpassing the baseline YOLO by 3.0% respectively. Furthermore, it achieved a detection speed of 305.0 Frame Per Second (FPS), demonstrating real-time capability. Ablation studies confirmed the individual contributions of FPSC, AFGC, and NWD modules, with the combined architecture showing synergistic improvements in both precision and computational efficiency. Comparative analyses against state-of-the-art attention mechanisms highlighted the superiority of AFGC in balancing parameter efficiency and detection accuracy. Ablation experiments confirmed that the combined architecture yielded the mAP gain and removing any one component produced a measurable drop in either precision, recall or localization stability.
In conclusion, the YOLO-FPSC algorithm provides a rapid, accurate, and computationally efficient solution for evaluating marine biofouling. By addressing key challenges in feature extraction, attention allocation, and loss optimization, this study advances the automation of biofouling assessment and offers a novel framework for evaluating antifouling coating performance. Future work will focus on extending the model to multi-species biofouling detection and integrating adaptive learning mechanisms for broader marine environmental monitoring applications.

Key words

deep learning / YOLO / biofouling evaluation / share convolution / attention mechanism / NWD

Cite this article

EndNote

Ris (Procite)

Bibtex

Download Citations

ZENG Chaofan, JIAO Jianbo, OU Zhihui, CHEN Baiyi, YANG Yanling, WU Jianhua. Evaluation Method of Marine Biofouling Based on Deep Learning[J]. Surface Technology. 2025, 54(24): 149-162

References

[1] 段继周, 刘超, 刘会莲, 等. 海洋水下设施生物污损及其控制技术研究进展[J]. 海洋科学, 2020, 44(8): 162-177.
DUAN J Z, LIU C, LIU H L, et al.Research Progress of Biofouling and Its Control Technology in Marine Underwater Facilities[J]. Marine Sciences, 2020, 44(8): 162-177.
[2] ALMEIDA E, DIAMANTINO T C, DE SOUSA O.Marine Paints: The Particular Case of Antifouling Paints[J]. Progress in Organic Coatings, 2007, 59(1): 2-20.
[3] CAO S, WANG J D, CHEN H S, et al.Progress of Marine Biofouling and Antifouling Technologies[J]. Chinese Science Bulletin, 2011, 56(7): 598-612.
[4] CAHILL P, TAIT L, FLOERL O, et al.A Portable Thermal System for Reactive Treatment of Biofouled Internal Pipework on Recreational Vessels[J]. Marine Pollution Bulletin, 2019, 139: 65-73.
[5] YANG W J, NEOH K G, KANG E T, et al.Polymer Brush Coatings for Combating Marine Biofouling[J]. Progress in Polymer Science, 2014, 39(5): 1017-1042.
[6] SEEBENS H, GASTNER M T, BLASIUS B.The Risk of Marine Bioinvasion Caused by Global Shipping[J]. Ecology Letters, 2013, 16(6): 782-790.
[7] YEBRA D M, KIIL S, DAM-JOHANSEN K.Antifouling Technology—Past, Present and Future Steps towards Efficient and Environmentally Friendly Antifouling Coatings[J]. Progress in Organic Coatings, 2004, 50(2): 75-104.
[8] 胡程颖, 张凯, 张陆, 等. 无锡自抛光防污涂料研究进展[J]. 中国涂料, 2024, 39(3): 24-29.
HU C Y, ZHANG K, ZHANG L, et al.Research Progress of Tin-Free Self-Polishing Antifouling Coatings[J]. China Coatings, 2024, 39(3): 24-29.
[9] CHEN L P, CUI R X, YAN W S, et al.Design and Climbing Control of an Underwater Robot for Ship Hull Cleaning[J]. Ocean Engineering, 2023, 274: 114024.
[10] DENG P C, WANG P L, HU J Z, et al.Design and Experimental Study of a Joint Protection System for Electrolytic Seawater Antifouling and Impressed Current Cathodic Protection[J]. Construction and Building Materials, 2025, 482: 141633.
[11] 黄璐琼, 谢胤, 冯娜. 舰船超声波防海生物污损技术应用试验研究[J]. 船舶工程, 2024, 46(9): 118-122.
HUANG L Q, XIE Y, FENG N.Experimental Study on Application of Ultrasonic Technology for Marine Organism Fouling Prevention on Ships[J]. Ship Engineering, 2024, 46(9): 118-122.
[12] MACLEOD N, BENFIELD M, CULVERHOUSE P.Time to Automate Identification[J]. Nature, 2010, 467(7312): 154-155.
[13] ZHAO L, ZHANG X L, YANG K D.A Deep Clustering Framework for Underwater Image Recognition[J]. Digital Signal Processing, 2025, 161: 105131.
[14] LECUN Y, BENGIO Y, HINTON G.Deep Learning[J]. Nature, 2015, 521(7553): 436-444.
[15] 张益. 基于卷积神经网络算法的海水循环冷却污损生物分类模型[J]. 工业水处理, 2024, 44(12): 160-165.
ZHANG Y.Convolution Neural Network Algorithm- Based Fouling Organisms Classification Model of Seawater Circulation Cooling System[J]. Industrial Water Treatment, 2024, 44(12): 160-165.
[16] CHIN C S, SI J T, CLARE A S, et al.Intelligent Image Recognition System for Marine Fouling Using Softmax Transfer Learning and Deep Convolutional Neural Networks[J]. Complexity, 2017, 2017(1): 5730419.
[17] GORMLEY K, MCLELLAN F, MCCABE C, et al.Automated Image Analysis of Offshore Infrastructure Marine Biofouling[J]. Journal of Marine Science and Engineering, 2018, 6(1): 2.
[18] MANNIX E J, WEI S S, A WOODHAM B, et al. Automating the Assessment of Biofouling in Images Using Expert Agreement as a Gold Standard[J]. Scientific Reports, 2021, 11: 2739.
[19] PEDERSEN M L, WEINELL C E, ULUSOY B, et al.Marine Biofouling Resistance Rating Using Image Analysis[J]. Journal of Coatings Technology and Research, 2022, 19(4): 1127-1138.
[20] MO C X, ZHU W Q, LU B Q, et al.Recognition Method of Turbine Pollutant Adhesion in Tidal Stream Energy Generation Systems Based on Deep Learning[J]. Energy, 2024, 302: 131799.
[21] RASHID H, HABBOUCHE H, AMIRAT Y, et al.B-FLOWS: Biofouling Focused Learning and Observation for Wide-Area Surveillance in Tidal Stream Turbines[J]. Journal of Marine Science and Engineering, 2024, 12(10): 1828.
[22] REDMON J, DIVVALA S, GIRSHICK R, et al.You Only Look Once: Unified, Real-Time Object Detection[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas, NV, USA. IEEE, 2016: 779-788.
[23] VARGHESE R, M S. YOLOv8: A Novel Object Detection Algorithm with Enhanced Performance and Robustness[C]//2024 International Conference on Advances in Data Engineering and Intelligent Computing Systems (ADICS). Chennai, India. IEEE, 2024: 1-6.
[24] IOFFE S, SZEGEDY C. Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift[EB/OL]. 2015: arXiv: 1502.03167. https:// arxiv.org/abs/1502.03167
[25] ELFWING S, UCHIBE E, DOYA K.Sigmoid-Weighted Linear Units for Neural Network Function Approximation in Reinforcement Learning[J]. Neural Networks, 2018, 107: 3-11.
[26] ZHAO B, CHONG Y S, DO A T.Area and Energy Efficient 2D Max-Pooling for Convolutional Neural Network Hardware Accelerator[C]//IECON 2020 The 46th Annual Conference of the IEEE Industrial Electronics Society. Singapore. IEEE, 2020: 423-427.
[27] LUO W, LI Y, URTASUN R, et al.Understanding the Effective Receptive Field in Deep Convolutional Neural Networks[C]//Proceedings of the 30th International Conference on Neural Information Processing Systems. Red Hook, NY, USA: Curran Associates Inc., 2016: 4905-4913.
[28] DIAKONIKOLAS I, KAMATH G, KANE D M, et al.Robustness Meets Algorithms[J]. Communications of the ACM, 2021, 64(5): 107-115.
[29] LIN T Y, DOLLÁR P, GIRSHICK R, et al. Feature Pyramid Networks for Object Detection[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu, HI, USA. IEEE, 2017: 936-944.
[30] LIU S, QI L, QIN H F, et al.Path Aggregation Network for Instance Segmentation[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, UT, USA. IEEE, 2018: 8759-8768.
[31] 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]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle, WA, USA. IEEE, 2020: 9756-9765.
[32] LI X, WANG W H, WU L J, et al. Generalized Focal Loss: Learning Qualified and Distributed Bounding Boxes for Dense Object Detection[EB/OL]. 2020: arXiv: 2006.04388. https://arxiv.org/abs/2006.04388
[33] ZHENG Z H, WANG P, REN D W, et al.Enhancing Geometric Factors in Model Learning and Inference for Object Detection and Instance Segmentation[J]. IEEE Transactions on Cybernetics, 2022, 52(8): 8574-8586.
[34] PIERRE S, ANDREA F, ESTEBAN R.Attention for Fine-Grained Categorization[C]//The Third International Conference on Learning Representations(ICLR). San Diego, CA, USA: Google, Inc., 2015: 1-11.
[35] XU C, WANG J W, YANG W, et al.Detecting Tiny Objects in Aerial Images: A Normalized Wasserstein Distance and a New Benchmark[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2022, 190: 79-93.
[36] LÉVY B, SCHWINDT E L. Notions of Optimal Transport Theory and how to Implement Them on a Computer[J]. Computers & Graphics, 2018, 72: 135-148.
[37] 曹正峰, 杜欣月, 李欣. 基于深度学习的再生骨料分类方法研究[J]. 建筑机械化, 2024, 45(11): 142-148.
CAO Z F, DU X Y, LI X.Research on Classification Method of Recycled Aggregate Based on Deep Learning[J]. Construction Mechanization, 2024, 45(11): 142-148.
[38] 王宁, 智敏. 深度学习下的单阶段通用目标检测算法研究综述[J]. 计算机科学与探索, 2025, 19(5): 1115-1140.
WANG N, ZHI M.Review of One-Stage Universal Object Detection Algorithms in Deep Learning[J]. Journal of Frontiers of Computer Science and Technology, 2025, 19(5): 1115-1140.
[39] WANG C Y, YEH I H, MARK LIAO H Y. YOLOv9: Learning What You Want to Learn Using Programmable Gradient Information[M]//Computer Vision - ECCV 2024. Cham: Springer Nature Switzerland, 2024: 1-21.
[40] CHEN H, CHEN K, DING G G, et al.YOLOv10: Real-Time End-to-End Object Detection[C]//Advances in Neural Information Processing Systems 37. Vancouver, BC, Canada. Neural Information Processing Systems Foundation, Inc.(NeurIPS), 2024: 107984-108011.

Funding

National Natural Science Foundation of China (22472067); Natural Science Foundation of Fujian Province (2024J01712); Fujian Provincial Transportation Science and Technology Program Project (YB202421); Xiamen Special Fund Project for Marine and Fishery Development (22CZB013HJ04)