Abstract
Fish population survey based on classification and recognition is an effective means to study water ecosystem. However, the identification of fish in the sea and other waters will be interfered by corals, reefs and other organisms. The great variety of fish also make it more difficult to distinguish. In order to improve the effect of fish recognition in the complex underwater environment, this paper proposes a method based on targeted sample transfer learning. The designed CNN is used to train the simple background fish data after background re-processing to obtain the pre-training model. By using transfer learning and interlayer fusion mechanism, the feature extraction layer of the pre-training model is frozen and fused with the new feature extraction layer in parallel, then, combined with pooling layer, a new feature extractor is formed, finally, connected to the classifier and output part to construct a new network, which is used to identify 10 kinds of fish with complex background. Compared with the original CNN model, the accuracy of the new network is improved by about 5%, reaching 91.33%. The experimental results show that the model can improve the ability of fish classification and recognition in complex underwater scenes, and can provide support for the study of fishery resources distribution.
Similar content being viewed by others
Data availability
All data and materials support our published claims and comply with field standards.
Code availability
All custom code support our published claims and comply with field standards.
References
Aa A et al (2020) Deep learning-based cross-machine health identification method for vacuum pumps with domain adaptation. Procedia Manuf 48:1088–1093
Albahli W, Albattah W (2020) Detection of coronavirus disease from X-ray images using deep learning and transfer learning algorithms. J X-Ray Sci Technol 28.5
Arthington AH et al (2016) Fish conservation in freshwater and marine realms: status, threats and management. Aquat Conserv: Mar Freshw Ecosyst 26(5):838–857
Moslem Azamfar, Xiang Li, Jay Lee (2020)Intelligent ball screw fault diagnosis using a deep domain adaptation methodology. Mechanism and Machine Theory,Volume 151,103932,ISSN 0094-114X. https://doi.org/10.1016/j.mechmachtheory.2020.103932
Boom BJ, Huang PX, He J, Fisher RB (2012) Supporting ground-truth annotation of image datasets using clustering, in: 2012 21st International Conference on Pattern Recognition (ICPR), IEEE, Tsukuba, Japan, pp 1542–1545
Chan A, Hodgson PA (2017) A systematic analysis of blast fishing in South-East Asia and possible solutions. 2017 IEEE Underwater Technology (UT). IEEE
Chan A, Hodgson PA (2019) A belt transect fish abundance survey methodology using an underwater vehicle. 2019 IEEE Underwater Technology (UT). IEEE
Chen W et al (2019) Fish classification based on deep convolutional neural network and transfer learning. Journal of Fuqing Branch of Fujian Normal University, No.5, Sum 156
Dai Y, Guojun W, Li K-C(2018) Conceptual alignment deep neural networks. J Intell Fuzzy Syst, 1631–1642
French B et al (2021) Comparing five methods for quantifying abundance and diversity of fish assemblages in seagrass habitat. Ecol Ind 124(4):107415
Hodgson G (2001) Reef Check: The first step in community-based management. Bull. Mar. Sci., 69, pp. 861-868
Huang Y, Wang Z (2020) Multi-granularity pruning for deep residual networks. J Intell Fuzzy Syst, 7403–7410. https://doi.org/10.3233/JIFS-200771
Li J et al (2021) Deep neural network-based real time fish detection method in the scene of marine fishing supervision. J Intell Fuzzy Syst, 1–6. https://doi.org/10.3233/JIFS-189713
Li J, Wu W, Xue D (2020) An intrusion detection method based on active transfer learning. Intell Data Anal, 363–383. https://doi.org/10.3233/IDA-194487
Lifei, Wang et al (2018) Modeling marine pelagic fish species spatiotemporal distributions utilizing a maximum entropy approach. Fish Oceanogr 27(6):571–586
Mostafa Mehdipour Ghazi, Berrin Yanikoglu, Erchan Aptoula (2017) Plant identification using deep neural networks via optimization of transfer learning parameters. Neurocomputing,Volume 235,Pages 228-235,ISSN 0925-2312. https://doi.org/10.1016/j.neucom.2017.01.018
Macias-Garcia E et al (2021) Multi-stage deep learning perception system for mobile robots. Integr Comput Aided Eng, 191–205. https://doi.org/10.3233/ICA-200640
Maruyama T et al (2018) Comparison of medical image classification accuracy among three machine learning methods. J X-Ray Sci Technol 26:1–93
Masoudi B, Daneshvar S, Razavi SN (2021) Multi-modal neuroimaging feature fusion via 3D Convolutional neural network architecture for schizophrenia diagnosis. Intell Data Anal, 527–540. https://doi.org/10.3233/IDA-205113
Pundhir S, Ghose U, Bisht U (2020) Assessment of effectiveness of data dependent activation method: MyAct. J Intell Fuzzy Syst, 665–677. https://doi.org/10.3233/JIFS-191618
Qin H et al (2016) Deep fish: accurate underwater live fish recognition with a deep architecture. Neurocomputing 187:49–58
Qiu C et al (2018) Transfer learning for small-scale fish image classification. 2018 OCEANS -. MTS/IEEE Kobe Techno-Ocean (OTO). IEEE
Salman A et al (2016) Fish species classification in unconstrained underwater environments based on deep learning. Limnology and Oceanography: Methods 14. https://doi.org/10.1002/lom3.10113
Yousaf, Waqas et al (2021) Patch-CNN: Deep learning for logo detection and brand recognition. J Intell Fuzzy Syst, 3849–3862. https://doi.org/10.3233/JIFS-190660
Xi, Qiao et al (2019) fvUnderwater sea cucumber identification based on Principal Component Analysis and Support Vector Machin. Measurement. https://doi.org/10.1016/j.measurement.2018.10.039
Yang F et al (2019) Quantification of hepatic steatosis in histologic images by deep learning method. J X-Ray Sci Technol, 1033–1045. https://doi.org/10.3233/XST-190570
Zheng Z, Fu J, Lu C et al (2021) Research on rolling bearing fault diagnosis of small dataset based on a new optimal transfer learning network. Measurement. https://doi.org/10.1016/j.measurement.2021.109285
M. D. Zeiler (2012) ADADELTA: An adaptive learning rate method. Computer ence, arXiv:1212.5701[cs.LG].
Author information
Authors and Affiliations
Contributions
All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Liangwei Jiang, Haiyan Quan, Tao Xie and Junbing Qian. The first draft of the manuscript was written by Liangwei Jiang and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.
Corresponding author
Ethics declarations
Conflicts of interest/Competing interests
The authors have no relevant financial or non-financial interests to disclose.
Ethics approval
The paper complies with ethical standards.
Consent to participate
All authors give consent to participate.
Consent for publication
All authors give consent for publication.
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Jiang, L., Quan, H., Xie, T. et al. Fish recognition in complex underwater scenes based on targeted sample transfer learning. Multimed Tools Appl 81, 25303–25317 (2022). https://doi.org/10.1007/s11042-022-12525-8
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11042-022-12525-8