Development and Evaluation of Advanced Morphological Algorithms for Automated Fish Measurement in Sustainable Fisheries

• Jáder Louis de Souza Gonçalves UNIR
• Carolina Rodrigues da Costa Dória UNIR
• Carolina Yukari Veludo Watanabe UNIR

Resumo

Accurate identification and measurement of fish populations present critical challenges in ecosystems characterized by significant morphological variability. This work presents morphological algorithms developed to enhance automated fish measurement processes in freshwater fisheries, with potential applications in coastal and marine ecosystems. Key contributions include developing algorithms tailored for non-linear morphologies and integrating these advancements into the ”ICTIOBIOMETRIA” application, which systematically records fish biometric data such as weight, dimensions, and capture locations. The dataset used originated from three primary sources: images inherited from prior collaborators, samples collected in partnership with the Ichthyology and Fisheries Laboratory at the Federal University of Rondônia, and acquisitions from fish markets. These sources provided 588 images representing diverse fish species from the Madeira River basin in Porto Velho, Rondônia, located in the Western Amazon. A dataset from established authors was also used as a reference to validate the proposed measurement models, ensuring a comprehensive comparative analysis that reinforced the robustness of the developed methods. To address challenges posed by larger fish and those exhibiting non-linear shapes, algorithms were developed and implemented to manage these characteristics effectively. Among the three measurement methods evaluated, the skeletonization method demonstrated superior performance, reducing the relative total error by 1.56%, and demonstrating robustness across diverse morphologies.

Referências

W. S. Rocha, T. F. C. da Fonseca, C. Y. V. Watanabe, C. R. da Costa Dória, and I. R. A. Sant’Anna, “Automatic measurement of fish from images using convolutional neural networks,” Multimedia Tools Appl., pp. 1–21, 2024, publisher: Springer.

B. Onsoy, A. Tarkan, H. Filiz, and G. Bilge, “Determination of the best length measurement of fish,” North-West. J. Zool., vol. 7, pp. 178–180, jun 2011.

M. Hao, H. Yu, and D. Li, “The measurement of fish size by machine vision—a review,” in Comput. Comput. Technol. Agric. IX: 9th IFIP WG 5.14 Int. Conf., CCTA 2015, Revised Selected Papers, Part II. Springer, 2016, pp. 15–32.

D. J. White, C. Svellingen, and N. J. C. Strachan, “Automated measurement of species and length of fish by computer vision,” Fish. Res., vol. 80, no. 2, pp. 203–210, 2006. [Online]. Available: [link]

N. Bravata, D. Kelly, J. Eickholt, J. Bryan, S. Miehls, and D. Zielinski, “Applications of deep convolutional neural networks to predict length, circumference, and weight from mostly dewatered images of fish,” Ecol. Evol., vol. 10, no. 17, pp. 9313–9325, 2020. [Online]. Available: [link]

A. Saleh, M. M. Hasan, H. W. Raadsma, M. S. Khatkar, D. R. Jerry, and M. Rahimi Azghadi, “Prawn morphometrics and weight estimation from images using deep learning for landmark localization,” Aquac. Eng., vol. 106, p. 102391, 2024. [Online]. Available: [link]

A. C. Müller and S. Guido, Introduction to Machine Learning with Python: A Guide for Data Scientists. Sebastopol, CA: O’Reilly Media, Inc., 2016.

W. McKinney, Python para análise de dados: Tratamento de dados com Pandas, NumPy e IPython. São Paulo: Novatec Editora, 2018.

N. N. C. Menezes, Introdução à programação com Python. São Paulo: Novatec Editora, 3rd ed., 2021.

T. H. Cormen, C. E. Leiserson, R. L. Rivest, and C. Stein, Introduction to Algorithms. Cambridge, MA: MIT Press, 2022.

L. Najman and H. Talbot, Mathematical Morphology: From Theory to Applications. London: Wiley-ISTE, 2013.

O. Lezoray and L. Grady, Image Processing and Analysis with Graphs: Theory and Practice. Boca Raton: CRC Press, 2012.