Detecção de alvos de peixes subaquáticos baseada em aprimoramento dinâmico de características

  • role: Corresponding author, First author通信作者, 第一作者
  • Affiliation:

    Yazhou Bay Innovation Institute, Hainan Tropical Ocean University, Sanya 572022, China

    College of Marine Science and Technology, Hainan Tropical Ocean University, Sanya 572022, China

  • Email:zhuxiaolonglong@sina.cn
  • Introduction:朱晓龙(1990-),男,黑龙江哈尔滨人,博士,讲师,硕士生导师,2013年、2016年和2020年于东北林业大学分别获得学士、硕士和博士学位,主要从事计算机视觉、深度学习及其在水下目标检测应用方面的研究。 E-mail:zhuxiaolonglong@sina.cn
ZHU Xiaolong12*,  
  • Affiliation:

    Yazhou Bay Innovation Institute, Hainan Tropical Ocean University, Sanya 572022, China

CHEN Yuwei1,  
  • Affiliation:

    Yazhou Bay Innovation Institute, Hainan Tropical Ocean University, Sanya 572022, China

WANG Jiayu1,  
  • Affiliation:

    Yazhou Bay Innovation Institute, Hainan Tropical Ocean University, Sanya 572022, China

GUO Haitao1,  
  • Affiliation:

    Yazhou Bay Innovation Institute, Hainan Tropical Ocean University, Sanya 572022, China

    College of Marine Science and Technology, Hainan Tropical Ocean University, Sanya 572022, China

CHEN Xiangzi12

Resumo

A monitorização eficiente de peixes subaquáticos é fundamental para a proteção dos ecossistemas marinhos, avaliação da biodiversidade e gestão sustentável dos recursos pesqueiros. Para enfrentar o impacto dos fatores ambientais complexos subaquáticos na estabilidade e eficiência da detecção, foi proposto um modelo dinâmico de aprimoramento de características FDN-YOLO (Fish Detection Network YOLO) baseado na estrutura YOLOv8n. Primeiramente, foi construído um módulo multi-escala de campo receptivo deformável (Multi-scale Deformable Receptive Field, MDRF), que ajusta adaptativamente o campo receptivo efetivo, permitindo que a rede principal represente melhor os alvos de peixes com diferentes formas e tamanhos. Em seguida, foi projetado um módulo leve de rearranjo espacial com redução de amostragem e separável em profundidade (Lite Space-to-Depth Depthwise Separable, Lite SPD-DS), que preserva pistas espaciais de alta granularidade na fase de redução de amostragem, controlando os custos computacionais. Por fim, foi proposta uma função de perda que combina IoU adaptativo com Varifocal Loss (Adaptive IoU-aware Varifocal Loss, AIVF Loss) para reforçar a capacidade de aprendizado de amostras de localização de alta qualidade e mitigar o viés causado pela distribuição desigual de classes e amostras. Os resultados experimentais baseados no conjunto de dados TF-DET mostram que o mAP50 e o mAP50∶95 do FDN-YOLO aumentaram 2,8% e 2,1%, respectivamente, enquanto a quantidade de parâmetros e o custo computacional foram reduzidos em 13,3% e 16,0%. Um experimento comparativo de generalização demonstrou ainda o excelente equilíbrio entre precisão, eficiência e estabilidade do FDN-YOLO, destacando seu potencial para aplicação em investigações ecológicas e gestão de recursos marinhos baseada em dados.

Palavras-chave

Detecção subaquática;Aprimoramento dinâmico;Aprendizado profundo;Redes neurais convolucionais;YOLOv8n

References

  1. 1.
    YAO T T, LI N, ZHANG Y. Perception enhanced hybrid network for underwater object detection[J]. Optics and Precision Engineering, 2025, 33(8): 1303-1312. (in Chinese). doi:
  2. 2.
    LI J, CONVERTINO M. Temperature increase drives critical slowing down of fish ecosystems[J]. PLoS One, 2021, 16(10): e0246222. doi:
  3. 3.
    POUR F B, PARRA L, LLORET J, et al. Measuring and evaluating the speed and the physical characteristics of fishes based on video processing[J]. Water, 2023, 15(11): 2138. doi:
  4. 4.
    HU J, ZHAO D D, ZHANG Y F, et al. Real-time nondestructive fish behavior detecting in mixed polyculture system using deep-learning and low-cost devices[J]. Expert Systems with Applications, 2021, 178: 115051. doi:
  5. 5.
    PENG Y F, ZHANG T Q, AN T. Channel adaptive underwater image enhancement algorithm based on pixel level[J]. Chinese Journal of Liquid Crystals and Displays, 2024, 39(8): 1037-1045. (in Chinese). doi:
  6. 6.
    LIU H Y, LIU T, GU Y Z, et al. A high-density fish school segmentation framework for biomass statistics in a deep-sea cage[J]. Ecological Informatics, 2021, 64: 101367. doi:
  7. 7.
    DONG C, WANG Z S, XING C D, et al. Occlusion detection of dense correspondence fields using motion statistics and triangulation[J]. IEEE Access, 2019, 7: 83569-83580. doi:
  8. 8.
    CARR E M, MOTTA P J. Tooth length and occlusion in four species of piscivorous fishes: getting a grip on prey[J]. Environmental Biology of Fishes, 2020, 103(8): 903-912. doi:
  9. 9.
    LI H Q, YU H, GAO H T, et al. Robust detection of farmed fish by fusing YOLOv5 with DCM and ATM[J]. Aquacultural Engineering, 2022, 99: 102301. doi:
  10. 10.
    MIAO L H, TIAN Y. Improved YOLOv8n multi-scale and lightweight underwater target detection[J]. Journal of Electronic Measurement and Instrumentation, 2025, 39(4): 141-151. (in Chinese)
  11. 11.
    LI J Y, LIU C N, LU X C, et al. CME-YOLOv5: an efficient object detection network for densely spaced fish and small targets[J]. Water, 2022, 14(15): 2412. doi:
  12. 12.
    ZHU J D, HE W W, WENG W D, et al. An embedding skeleton for fish detection and marine organisms recognition[J]. Symmetry, 2022, 14(6): 1082. doi:
  13. 13.
    LU Y C, TUNG C, KUO Y F. Identifying the species of harvested tuna and billfish using deep convolutional neural networks[J]. ICES Journal of Marine Science, 2020, 77(4): 1318-1329. doi:
  14. 14.
    XU X L, LI W S, DUAN Q L. Transfer learning and SE-ResNet152 networks-based for small-scale unbalanced fish species identification[J]. Computers and Electronics in Agriculture, 2021, 180: 105878. doi:
  15. 15.
    IQBAL M A, WANG Z J, ALI Z A, et al. Automatic fish species classification using deep convolutional neural networks[J]. Wireless Personal Communications, 2021, 116(2): 1043-1053. doi:
  16. 16.
    SMADI AAL, MEHMOOD A, ABUGABAH A, et al. Deep convolutional neural network-based system for fish classification[J]. International Journal of Electrical and Computer Engineering (IJECE), 2022, 12(2): 2026. doi:
  17. 17.
    SAYDIRASULOVICH S N, MUKHIDDINOV M, DJURAEV O, et al. An improved wildfire smoke detection based on YOLOv8 and UAV images[J]. Sensors, 2023, 23(20): 8374. doi:
  18. 18.
    ZHANG B, ZHU Z Y, XU J Z. Deep-sea pollutant detection for autonomous underwater robots[J]. Optics and Precision Engineering, 2025, 33(11): 1803-1817. (in Chinese)
  19. 19.
    SONG D K, HUO H, SONG D K, et al. Lightweight underwater target detection algorithm based on YOLOv8n[J]. Electronics, 2025, 14(9): 1810. doi:
  20. 20.
    DAI J F, QI H Z, XIONG Y W, et al. Deformable convolutional networks[C]. 2017 IEEE International Conference on Computer Vision (ICCV).22-29,2017, Venice, Italy. IEEE, 2017: 764-773. doi:
  21. 21.
    XIONG Y W, LI Z Q, CHEN Y T, et al. Efficient deformable ConvNets: rethinking dynamic and sparse operator for vision applications[C]. 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).16-22,2024, Seattle, WA, USA. IEEE, 2024: 5652-5661. doi:
  22. 22.
    SUNKARA R, LUO T. No more strided convolutions orPooling: a new CNN building block forLow-Resolution images andSmall objects[C]. Machine Learning and Knowledge Discovery in Databases. Cham: Springer, 2023: 443-459. doi:
  23. 23.
    HOWARD A G, ZHU M L, CHEN B, et al. MobileNets: efficient convolutional neural networks for mobile vision applications[EB/OL]. 2017: arXiv: 1704.04861. .
  24. 24.
    ZHANG H Y, WANG Y, DAYOUB F, et al. VarifocalNet: an IoU-aware dense object detector[C]. 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).20-25,2021, Nashville, TN, USA. IEEE, 2021: 8510-8519. doi:
  25. 25.
    LU Z X, ZHU X L, GUO H T, et al. FishFocusNet: an improved method based on YOLOv8 for underwater tropical fish identification[J]. IET Image Processing, 2024, 18(12): 3634-3649. doi:
  26. 26.
    WANG C Y, BOCHKOVSKIY A, LIAO H M. 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).17-24,2023, Vancouver, BC, Canada. IEEE, 2023: 7464-7475. doi:
  27. 27.
    SOHAN M, SAI RAM T, RAMI REDDY C V. A review on YOLOv8 and its advancements[C]. Data Intelligence and Cognitive Informatics. Singapore: Springer, 2024: 529-545. doi:
  28. 28.
    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. December 10-15, 2024. Vancouver, BC, Canada. Neural Information Processing Systems Foundation, Inc. (NeurIPS), 2024: 107984-108011.
  29. 29.
    TIAN Y J, YE Q X, DOERMANN D. YOLOv12: attention-centric real-time object detectors[EB/OL]. 2025: arXiv: 2502.12524. 。
  30. 30.
    SHEN X, YUAN G L, WANG H B, et al. Unsupervised clustering optimization-based efficient attention in YOLO for underwater object detection[J]. Artificial Intelligence Review, 2025, 58(7): 219. doi:
  31. 31.
    FU C P, LIU R S, FAN X, et al. Rethinking general underwater object detection: Datasets, challenges, and solutions[J]. Neurocomputing, 2023, 517(C): 243-256. doi:
  32. 32.
    LIU C W, LI H J, WANG S C, et al. A dataset and benchmark of underwater object detection for robot picking[C]. 2021 IEEE International Conference on Multimedia & Expo Workshops (ICMEW).5-9,2021, Shenzhen, China. IEEE, 2021: 1-6. doi:

Leia o texto completo

The above content is generated by Large Model Translation. The translated content is for reference only. We do not assume any commercial or legal responsibilty for any consequences arising from the use of our website