[論文レビュー] Survey of Automatic Plankton Image Recognition: Challenges, Existing Solutions and Future Perspectives

Tuomas Eerola, Daniel Batrakhanov+8 more|arXiv (Cornell University)|May 19, 2023

Water Quality Monitoring Technologies被引用数 8

ひとこと要約

A comprehensive survey of automatic plankton image recognition, detailing challenges, current solutions (feature engineering and CNN-based methods), datasets, and future directions towards instrument- and location-agnostic systems.

ABSTRACT

Planktonic organisms are key components of aquatic ecosystems and respond quickly to changes in the environment, therefore their monitoring is vital to understand the changes in the environment. Yet, monitoring plankton at appropriate scales still remains a challenge, limiting our understanding of functioning of aquatic systems and their response to changes. Modern plankton imaging instruments can be utilized to sample at high frequencies, enabling novel possibilities to study plankton populations. However, manual analysis of the data is costly, time consuming and expert based, making such approach unsuitable for large-scale application and urging for automatic solutions. The key problem related to the utilization of plankton datasets through image analysis is plankton recognition. Despite the large amount of research done, automatic methods have not been widely adopted for operational use. In this paper, a comprehensive survey on existing solutions for automatic plankton recognition is presented. First, we identify the most notable challenges that that make the development of plankton recognition systems difficult. Then, we provide a detailed description of solutions for these challenges proposed in plankton recognition literature. Finally, we propose a workflow to identify the specific challenges in new datasets and the recommended approaches to address them. For many of the challenges, applicable solutions exist. However, important challenges remain unsolved: 1) the domain shift between the datasets hindering the development of a general plankton recognition system that would work across different imaging instruments, 2) the difficulty to identify and process the images of previously unseen classes, and 3) the uncertainty in expert annotations that affects the training of the machine learning models for recognition. These challenges should be addressed in the future research.

研究の動機と目的

• Identify and categorize the key challenges hindering automatic plankton recognition for large-scale operational use.
• Review existing methods (feature engineering and deep learning) and their applicability to plankton images.
• Summarize public datasets and imaging instruments used in plankton recognition research.
• Propose a workflow to diagnose dataset-specific challenges and recommended remedies.
• Highlight unsolved challenges and future research directions for harmonized, instrument- and location-agnostic solutions.

提案手法

• Survey literature to identify major challenges in plankton recognition and summarize corresponding solutions from the literature.
• Describe traditional feature-engineering based pipelines and their typical components (feature extraction, selection, classification).
• Review CNN-based architectures, including general-purpose and custom plankton-specific models, and analyze their performance trends across datasets.
• Discuss hybrid approaches combining handcrafted features with CNNs and ensemble strategies.
• Summarize publicly available plankton image datasets and imaging instruments to contextualize dataset differences.

実験結果

リサーチクエスチョン

• RQ1 What are the primary challenges blocking general, operational plankton recognition across different imaging systems?
• RQ2 What solutions exist in the literature to address domain shift, unseen classes, and annotation uncertainty in plankton recognition?
• RQ3 How do feature-engineering methods compare with CNN-based approaches for plankton image classification across benchmarks?
• RQ4 What datasets and imaging instruments are available, and how do they influence method performance and comparability?
• RQ5 What workflow can guide addressing dataset-specific challenges in new plankton image collections?

主な発見

• Domain shift between datasets hampers development of a general plankton recognition system usable across instruments.
• Identifying and processing images of unseen classes and non-plankton particles remains difficult.
• Uncertainty in expert annotations affects training of recognition models.
• There exist applicable solutions for many challenges, but some important challenges remain unsolved, particularly for operational deployment.
• CNN-based methods have largely surpassed traditional feature-engineering approaches in accuracy on several benchmarks, yet operational adoption is limited.
• Hybrid and ensemble approaches, as well as tailored architectures for holographic and specialized imaging, are explored to improve performance and efficiency.