Automated identification of copepods using digital image processing and artificial neural network

Area of Science:

• Marine Biology
• Computational Biology
• Ecology

Background:

• Copepods are crucial planktonic organisms in marine food webs, influencing trophodynamics and fisheries.
• Accurate identification of copepod species is vital for ecological assessments but is traditionally labor-intensive and time-consuming.
• Automated methods are needed to streamline copepod identification and classification.

Purpose of the Study:

• To develop and evaluate an automated technique for identifying and classifying copepod specimens.
• To apply digital image processing and machine learning for efficient copepod analysis.
• To assess the system's accuracy across multiple copepod species.

Main Methods:

• Extracted morphological features from copepod images using digital image processing.
• Employed an Artificial Neural Network (ANN) for species classification.
• Utilized a dataset with 60% for training and 40% for testing the classification model.

Main Results:

• Achieved an overall classification accuracy of 93.13% for eight copepod species.
• Demonstrated high accuracy for specific species: 100% for Acartia spinicauda, Bestiolina similis, and Oithona aruensis.
• Showcased varying but high accuracies for other species, including Oithona dissimilis (90%) and Tortanus barbatus (95%).

Conclusions:

• The developed method enables rapid, automated species-level classification of copepods.
• Future work should expand the model to include more species and refine feature selection.
• Reducing image capture time is a potential area for future optimization.