Related Experiment Video
Updated: Jun 21, 2025
10:56
Long-term Behavioral Tracking of Freely Swimming Weakly Electric Fish
Published on: March 6, 2014
12.5K
Non-Invasive Fish Biometrics for Enhancing Precision and Understanding of Aquaculture Farming through Statistical
Fernando Joaquín Ramírez-Coronel1, Oscar Mario Rodríguez-Elías1, Edgard Esquer-Miranda2
1Division of Graduate Studies and Research, Tecnológico Nacional de México/ I.T. de Hermosillo, Av. Tecnológico 115, Hermosillo 83170, Sonora, Mexico.
Animals : an Open Access Journal From MDPI
|July 13, 2024
Summary
This study introduces a computer vision method for accurate, non-invasive fish biomass estimation in aquaculture. The approach uses machine learning and statistical analysis of fish size and shape, showing robust performance for tilapia.
Area of Science:
- Aquaculture and Aquatic Resources
- Computer Vision and Machine Learning
- Fish Biology and Ecology
Background:
- Accurate fish biomass estimation is crucial for effective aquaculture management.
- Current non-invasive methods often lack precision or require manual intervention.
- Developing automated, reliable fish biometrics is a key challenge in the industry.
Purpose of the Study:
- To validate a novel computer vision methodology for precise, non-invasive fish biomass estimation.
- To assess the performance of signature function-based feature extraction combined with machine learning.
- To compare automatically extracted features against manually extracted ones under varying conditions.
Main Methods:
- Utilized a signature function-based feature extraction algorithm for statistical morphological analysis.
- Applied three common machine learning methods (including multilayer perceptron) to tilapia (Oreochromis niloticus) images.
- Evaluated performance under two different lighting conditions using a dataset of 129 tilapia samples.
Main Results:
- The multilayer perceptron model demonstrated robust and superior accuracy across different features and lighting conditions.
- Automatically extracted features showed competitive results compared to manually extracted ones.
- The model's interpretability offers insights into fish morphological and allometric changes.
Conclusions:
- The proposed computer vision methodology significantly advances non-invasive fish biometrics for aquaculture.
- The approach offers improved precision, interpretability, and versatility across species and developmental stages.
- This research paves the way for real-time fish welfare monitoring and biomass estimation in fish farms.
