Open AccessEstimating reef fish size distributions with a mini remotely operated vehicle-integrated stereo camera system
Author(s) -
Steven B. Garner,
Aaron M. Olsen,
Ryan Caillouet,
Matthew D. Campbell,
William F. Patterson
Publication year - 2021
Publication title -
plos one
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.99
H-Index - 332
ISSN - 1932-6203
DOI - 10.1371/journal.pone.0247985
Subject(s) - fish measurement , remotely operated vehicle , reef , remotely operated underwater vehicle , environmental science , coral reef fish , range (aeronautics) , fish <actinopterygii> , remote sensing , statistics , geology , mathematics , fishery , biology , computer science , materials science , oceanography , artificial intelligence , robot , composite material , mobile robot
We tested the efficacy of a stereo camera (SC) system adapted for use with a remotely operated vehicle (ROV) to estimate fish length distributions at reef sites in the northern Gulf of Mexico. A pool experiment was conducted to test the effect of distance (1, 2, 3 or 5 m), angle of incidence (AOI; 0° to 40° at 5° increments), and SC baseline distance (BD; BD1 = 406, BD2 = 610, and BD3 = 762 mm camera separation) on the accuracy and precision of fish model length (288, 552, or 890 mm fork length) estimates compared to a red laser scaler (RLS). A field experiment was then conducted at 20 reef sites with SCs positioned at BD1 to compare fish length distribution estimates between the SC and RLS systems under in situ conditions. In the pool experiment, mean percent errors were consistently within the a priori selected threshold of ±5% at AOIs ≤10° at all distances with all four systems. However, SCs produced accurate estimates at AOIs up to 30° at all distances tested; 2–3 m was optimal. During reef site surveys, SCs collected 10.4 times as many length estimates from 4.3 times as many species compared to the RLS. Study results demonstrate that, compared to laser scalers, ROV-based SC systems can substantially increase the number of available fish length estimates by producing accurate length estimates at a wider range of target orientations while also enabling measurements from a greater portion of the cameras’ field of view.