Classifying Sonar Images: Can a Computer-Driven Process Identify Eels?
Bibliographic record
Abstract
Abstract The purpose of this study was to explore the extent to which a computer-driven process can be used to classify sonar images. The data we present come from a feasibility study for a hydroacoustic monitoring system aimed at the automatic detection of downstream-migrating adult American eels Anguilla rostrata in the intake canal of a small hydroelectric station. The images were collected by a dual-frequency identification sonar with sufficient resolution to show the distinct shape and swimming motion of eels, and thus to allow confident visual identification. The goal was to find a set of image processing, tracking, and pattern recognition techniques that would reproduce the results of the visual classification. Of the three classification methods that we tested with our example data set, neural network analysis had the lowest misclassification rate for eels (7% of the eels being misclassified as debris) and the second-lowest misclassification rate for debris (5% of the debris being misclassified as eels). Discriminant function analysis misclassified 12% of the eels as debris and 4% of the debris as eels. A K-nearest-neighbor analysis initially provided the poorest results (17% misclassified eels and 12% misclassified debris). However, after applying an algebraic correction, K-nearest-neighbor analysis yielded an accurate estimate of the number of eels in the data set. We discuss the value of flagging cases of uncertain classification, how image processing and feature selection can affect the results, and how the numeric ratio of the targets present determines what error rates are acceptable. We conclude that, depending on the application, different degrees of automation may be achieved, ranging from a relatively high degree of human supervision in the classification of all potential targets to a fully automated process that requires only periodic quality control and adjustments of the classification model.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.004 | 0.025 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.002 | 0.001 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.001 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".