Computer Vision Advances in Monitoring Multi‐Species Invertebrate Fisheries in the Pacific

ABSTRACT Monitoring diverse, dispersed small‐scale invertebrate fisheries remains challenging because taxonomic diversity, informal trade and dispersed harvesting limit sustained programs that can capture standardised data used for catch identification and measurement. Computer vision is increasingly being used to overcome these challenges, but most affordable and accessible single‐camera systems face a trade‐off: close‐range imagery required for high fidelity species identification while increasing parallax error for three‐dimensional specimens. We tested a hardware‐agnostic calibration method to improve an automated fisheries‐monitoring platform. Nationally standardised coins placed on specimens provided a local, distance‐independent scale reference. Manual, coin‐corrected and uncorrected computer‐vision measurements of carapace length in spiny lobsters ( Panulirus penicillatus ) were compared using linear mixed‐effects models. Coin correction removed systematic bias, reducing mean percentage error from 10.77% to 1.20% and mean absolute error from 8.6 to 2.3 mm. By contrast, linear‐regression correction for Panulirus penicillatus under‐corrected small lobsters and over‐corrected large individuals, biasing Length‐Based Spawning Potential Ratio estimates. Where remote observers could verify species‐level labels, computer‐vision classification exceeded 90% accuracy, with other performance metrics scoring above 80%. Performance on unfiltered operational data depended on image acquisition: close‐range board imagery substantially outperformed greater‐distance mat imagery for taxa requiring fine diagnostic detail, particularly bivalves and gastropods. Coplanar calibration is therefore necessary for 2D computer‐vision systems to provide reliable species identification and morphometric data simultaneously, supporting scalable, field‐appropriate monitoring of multi‐species invertebrate fisheries in data‐limited small‐island developing states.

Authors

Institutions

Publication Details

Journal
Fish and Fisheries
Published
2026-11-13
DOI
https://doi.org/10.1111/faf.70122
Primary Topic
Water Quality Monitoring Technologies
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Computer Vision Advances in Monitoring Multi‐Species Invertebrate Fisheries in the Pacific

F. Magron, Tim Adams, J. Aiafi, Colette C. C. Wabnitz et al.
Fish and Fisheries
Water Quality Monitoring Technologies
article

Computer Vision Advances in Monitoring Multi‐Species Invertebrate Fisheries in the Pacific

F. Magron, Tim Adams, J. Aiafi, Colette C. C. Wabnitz, George Shedrawi, Pauline Bosserelle, Dirk J. Steenbergen, Owen Li, Ulusapeti Tiitii, Sebastien Gislard, Viliami Fatongiatau, Chris Molai, Shivam Jalam, Faasulu Fepuleai, Christina Mosile, Meliki Rokuro, Jefferey P. Kinch, Hannah Gilchrist, Isaako Papalii, Prakriti Rachna, Neil L. Andrew, Leonore Page, Diana Divalotu
article en

Abstract

ABSTRACT Monitoring diverse, dispersed small‐scale invertebrate fisheries remains challenging because taxonomic diversity, informal trade and dispersed harvesting limit sustained programs that can capture standardised data used for catch identification and measurement. Computer vision is increasingly being used to overcome these challenges, but most affordable and accessible single‐camera systems face a trade‐off: close‐range imagery required for high fidelity species identification while increasing parallax error for three‐dimensional specimens. We tested a hardware‐agnostic calibration method to improve an automated fisheries‐monitoring platform. Nationally standardised coins placed on specimens provided a local, distance‐independent scale reference. Manual, coin‐corrected and uncorrected computer‐vision measurements of carapace length in spiny lobsters ( Panulirus penicillatus ) were compared using linear mixed‐effects models. Coin correction removed systematic bias, reducing mean percentage error from 10.77% to 1.20% and mean absolute error from 8.6 to 2.3 mm. By contrast, linear‐regression correction for Panulirus penicillatus under‐corrected small lobsters and over‐corrected large individuals, biasing Length‐Based Spawning Potential Ratio estimates. Where remote observers could verify species‐level labels, computer‐vision classification exceeded 90% accuracy, with other performance metrics scoring above 80%. Performance on unfiltered operational data depended on image acquisition: close‐range board imagery substantially outperformed greater‐distance mat imagery for taxa requiring fine diagnostic detail, particularly bivalves and gastropods. Coplanar calibration is therefore necessary for 2D computer‐vision systems to provide reliable species identification and morphometric data simultaneously, supporting scalable, field‐appropriate monitoring of multi‐species invertebrate fisheries in data‐limited small‐island developing states.

Fish and Fisheries
Pacific Community (NC), Stockholm Resilience Centre (SE), University of British Columbia (CA), Stockholm University (SE), University of Wollongong (AU), Ministry of Agriculture and Fisheries (WS), Ministry of Agriculture and Forestry and Food (TO), Institut Louis Malardé (PF), University of the South Pacific (FJ), Stanford University (US)
Water Quality Monitoring Technologies
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon