Automated Classification of Marine Plankton Images

Marine plankton form the foundation of oceanic food webs and play a pivotal role in global biogeochemical cycles. Automated classification of plankton images harnesses advances in digital optics, imaging flow cytometry and in situ microscopy to replace labour-intensive manual sorting. By applying sophisticated algorithms to digitised images, researchers can rapidly identify and count diverse plankton taxa across vast spatial and temporal scales. This capability is essential for monitoring ecosystem health, detecting harmful algal blooms and informing fisheries management.

State-of-the-art approaches employ machine learning and deep learning models—especially convolutional neural networks—to extract morphological and textural features directly from raw pixel data. These methods overcome many limitations of traditional image-processing pipelines, such as rigid feature-engineering and sensitivity to background noise. Key challenges include heterogeneity in imaging instruments, variability in light conditions, the occurrence of previously unseen species and the scarcity of expertly annotated training data. Ongoing research seeks to improve generalisation across devices (domain-agnostic models), to quantify uncertainty in predictions and to integrate classification outputs into real-time monitoring platforms.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Automated Classification of Marine Plankton Images publication trend

The graph below shows the total number of articles in automated classification of marine plankton images across all publications each year (not limited to Nature Index journals).

Technical terms

Deep learning: A class of machine learning techniques using layered neural networks to learn hierarchical representations from image data for classification tasks.

Domain shift: Variability between training and new datasets or imaging instruments that can reduce model generalisation.

Mean average precision (mAP): A summary metric evaluating detection performance by averaging precision across multiple recall thresholds.

Functional traits: Morphological or behavioural characteristics of plankton that can be quantitatively extracted from image data.

Tracking accuracy: The proportion of correctly associated individuals across consecutive image frames in an automated tracking system.

References

1. Survey of automatic plankton image recognition: challenges, existing solutions and future perspectives. Artificial Intelligence Review (2024).
2. Rotifer detection and tracking framework using deep learning for automatic culture systems. Smart Agricultural Technology (2024).
3. Machine learning techniques to characterize functional traits of plankton from image data. Limnology and Oceanography (2022).