Page 249 - 《水产学报》2026年第3期
P. 249

3 期                                     水    产    学    报                                 50 卷






                  20 mm     1  25 mm     2  20 mm    3  20 mm    4  50 mm    5  42 mm    6  30 mm     7






                  20 mm     8  25 mm    9  20 mm    10  20 mm    11  50 mm   12  42 mm   13  30 mm   14







                  20 mm     15  25 mm      20 mm    17  20 mm    18  50 mm      42 mm    20  30 mm   21
                                        16                                   19
                                        图版 Ⅲ    YOLOv11-AMSTAR   与不同模型检测对比
              1~7.原始图像,8~14. YOLOv11  的检测结果,15~21. YOLOv11-AMSTAR  的检测结果。
                       Plate Ⅲ Comparison of YOLOv11-AMSTAR and detection performance with different models
              1-7. original image, 8-14. the detection results of YOLOv11, 15-21. the detection results of YOLOv11-AMSTAR.
              场在线实时性未能得到很好训练。②模型的泛化                           [  4  ]   Kibet D, Shin J H. Counting abalone with high precision using
              能力未得到验证,由于数据集样本种类与数量限                                 YOLOv3 and DeepSORT[J]. Processes, 2023, 11(8): 2351.
              制,未能进一步将模型推广到其他海珍类如泥东                           [  5  ]   Wang H H, Xing S Y, Zheng Y S, et al. Study on the shape
              风 螺  (Babylonia  lutosa)、 栉 孔 扇 贝  (Chlamys  far-     detection method for the precious seafoods based on computer
              reri) 等高经济价值小目标监测任务。所以未来的                             vision[J]. Agricultural Engineering, 2015, 47(3): 113-120.
              研究可以聚焦于水下机器人捕捞移动目标鲍的实                           [  6  ]   Li J, Zhang Y, Liu S, et al. Insights into adhesion of abalone: a
              际场景,进一步改善模型对移动目标的识别能力,                                mechanical approach[J]. Journal of the Mechanical Behavior of
              提高模型的泛化能力,以适应更复杂的水下多目                                 Biomedical Materials, 2018, 77: 331-336.
              标、多类别实时检测需求,为智慧渔业提供更加                           [  7  ]   Chen Y H, Lin D Y, Lan W Y, et al. Underwater measurement
              智能、可靠和高效的解决方案。                                        of abalone shell’s size based on stereo vision[J]. Measurement,
                                                                    2025, 242: 115699.
              (作者声明本文无利益冲突)                                   [  8  ]   Han M, Lyu Z Y, Qiu T, et al. A review on intelligence dehaz-
                                                                    ing and color restoration for underwater images[J]. IEEE Trans-
              参考文献     (References):                                actions  on  Systems,  Man,  and  Cybernetics:  Systems,  2020,
              [  1  ]   Huang  H,  Tang  Q  R,  Li  J  Y,  et  al.  A  review  on  underwater  50(5): 1820-1832.
                    autonomous  environmental  perception  and  target  grasp,  the  [  9  ]   Zhang T, Su H B, Fan B, et al. Underwater image enhancement
                    challenge  of  robotic  organism  capture[J].  Ocean  Engineering,  based  on  red  channel  correction  and  improved  multiscale
                    2020, 195: 106644.                              fusion[J]. IEEE Transactions on Geoscience and Remote Sens-
              [  2  ]   Ye M G, Li J. Abalone counting based on improved Faster R-  ing, 2024, 62: 4205120.
                    CNN[C]//Association  for  Computing  Machinery.  Proceedings  [10]   Yan S Z, Chen X Y, Wu Z X, et al. HybrUR: a hybrid physical-
                    of the 2022 2nd International Conference on Bioinformatics and  neural   solution   for   unsupervised   underwater   image
                    Intelligent  Computing.  Harbin:  Association  for  Computing  restoration[J].  IEEE  Transactions  on  Image  Processing,  2023,
                    Machinery, 2022: 206-210.                       32: 5004-5016.
              [  3  ]   Park K M, Ahn B W, Park Y S, et al. A study on abalone young  [11]   Kang Q Z, Ye F, Li Q, et al. Bubble boundary R-CNN: a multi-
                    shells counting system using machine vision[J]. Journal of the  task  model  for  segmenting  and  reconstructing  overlapping
                    Korean  Society  of  Marine  Environment  and  Safety,  2017,  bubbles[J]. Separation and Purification Technology, 2025, 358:
                    23(4): 415-420.                                 130300.

              https://www.china-fishery.cn                           中国水产学会主办    sponsored by China Society of Fisheries
                                                            10
   244   245   246   247   248   249   250   251   252   253   254