Page 37 - 《水产学报》2025年第7期
P. 37
张东,等 水产学报, 2025, 49(7): 079103
工作的重点之一。 Society B: Biological Sciences, 2010, 365(1550): 2163-2176.
3.7 数据质量与标准化 [ 2 ] Matley J K, Klinard N V, Barbosa Martins A P, et al. Global
trends in aquatic animal tracking with acoustic telemetry[J].
了解动物的行为对于行为学知识的广泛应
Trends in Ecology & Evolution, 2022, 37(1): 79-94.
用具有重要意义。然而,现有的动物行为数据
[ 3 ] Yang C, Su X P, Liu D P, et al. A new method of aquatic
集在多个方面存在局限性,包括动物类别、数
animal personality analysis based on machine learning
据样本和提供的任务数量有限,以及环境条件
(PAML): taking swimming crab Portunus trituberculatus as an
和视角的变化有限。为了更全面地了解自然动
example[J]. Frontiers in Marine Science, 2020, 7: 32.
物行为,数据集应该涵盖广泛的环境,包括背
[ 4 ] Arablouei R, Wang Z W, Bishop-Hurley G J, et al. Multimodal
景、视角、光照和天气条件的变化。
sensor data fusion for in-situ classification of animal behavior
采集海量数据为 AI 模型提供训练素材固
using accelerometry and GNSS data[J]. Smart Agricultural
然重要,但高质量、格式统一的数据更重要。
Technology, 2023, 4: 100163.
高质量数据能让 AI 模型准确学习模式和规律,
[ 5 ] Banerjee S C, Khan K A, Sharma R. Deep-worm-tracker: deep
做出正确决策,并提高 AI 模型的泛化能力。未
learning methods for accurate detection and tracking for behavi-
来需要制定统一的水生动物行为学数据采集与
标准,建立共享平台,以便于跨机构、跨区域 oral studies in C. elegans[J]. Applied Animal Behaviour Sci-
ence, 2023, 266: 106024.
的数据整合和比较。
[ 6 ] Lim L W K. Implementation of artificial intelligence in
3.8 跨学科协作与应用推广
aquaculture and fisheries: deep learning, machine vision, big
水生动物行为学研究涉及生物学、生态学、 data, internet of Things, robots and beyond[J]. Journal of Com-
计算机科学和工程学等多学科,未来需要加强 putational and Cognitive Engineering, 2024, 3(2): 112-118.
跨学科合作,推动研究成果向实际应用转化。 [ 7 ] Saad Saoud L, Sultan A, Elmezain M, et al. Beyond observa-
同时,应注重提升公众和管理者对 AI 技术的认 tion: deep learning for animal behavior and ecological conser-
识和接受度,促进科技成果服务于水域生态保 vation[J]. Ecological Informatics, 2024, 84: 102893.
护和水产业的可持续发展。 [ 8 ] Zhao Y X, Qin H X, Xu L, et al. A review of deep learning-
based stereo vision techniques for phenotype feature and beha-
4 结论
vioral analysis of fish in aquaculture[J]. Artificial Intelligence
不可否认,AI 正在引领动物行为研究的新 Review, 2025, 58(1): 7.
时代,为水生动物行为学研究带来了前所未有 [ 9 ] Cui M, Liu X B, Liu H H, et al. Fish tracking, counting, and
的机遇,AI 正将水生动物行为学研究推向“数 behaviour analysis in digital aquaculture: a comprehensive sur-
据密集型科学”新范式。自动化图像处理、行为 vey[J]. Reviews in Aquaculture, 2025, 17(1): e13001.
模式识别和实时监控系统正逐步改变传统研究 [10] Boudhane M, Nsiri B. Underwater image processing method for
方式,使得数据采集和分析更高效、准确,从 fish localization and detection in submarine environment[J].
而进一步揭示隐藏的模式,更好地理解复杂的 Journal of Visual Communication and Image Representation,
行为,推动对自然世界的深入了解。尽管面临 2016, 39: 226-238.
数据标准化、模型泛化和跨学科协作等挑战, [11] Spampinato C, Giordano D, Di Salvo R, et al. Automatic fish
未来随着技术的不断进步与应用推广,AI 将在 classification for underwater species behavior understanding
水生动物行为学研究中发挥更加重要的作用, [C]//Association for Computing Machinery. Proceedings of the
助力水域生态保护与水产业的可持续发展。 First ACM International Workshop on Analysis and Retrieval
of Tracked Events and Motion in Imagery Streams. Firenze
参考文献 (References): Italy: Association for Computing Machinery, 2010: 45-50.
[ 1 ] Tomkiewicz S M, Fuller M R, Kie J G, et al. Global position- [12] Zheng T, Wu J F, Kong H, et al. A video object segmentation-
ing system and associated technologies in animal behaviour and based fish individual recognition method for underwater com-
ecological research[J]. Philosophical Transactions of the Royal plex environments[J]. Ecological Informatics, 2024, 82: 102689.
中国水产学会主办 sponsored by China Society of Fisheries https://www.china-fishery.cn
5