Somerton et al.: Quantifying the behavior of fish in response to a moving camera vehicle 
345 
Figure 1 
Photographs of (A) an observation cage, containing the Modular Optical Underwater Survey System, a DIDSON imaging 
sonar (black rectangular box), and the bridles used to attach the cage to the longline (yellow lines at the top), and (B) the 
camera-based assessment survey system (C-BASS) towed camera vehicle shown before being launched from the stern of 
the RV Pelican. The cage and camera vehicle were used in an experiment conducted in July 2014 in the northeast Gulf of 
Mexico. 
an important commercial and recreational species and 
is a dominant component of the reef fish assemblage 
in the southeastern United States. During the night, 
vermilion snapper forage on planktonic and benthic 
prey on both reefs and their adjacent sand flats. How- 
ever, during the day, they are reported to form resting 
schools on the tops of reefs (Grimes, 1979; Sedberry 
and Cuellar, 1993). Viewed from our cameras, the be- 
havior of the school of vermilion snapper in the study 
area was consistent with this notion. 
To quantify changes in fish behavior, an observa- 
tional test bed was set up at select locations on a daily 
basis. The test bed consisted of 3 observation cages 
(Fig. 1) measuring 94 cm wide, 94 cm long, and 67 cm 
high. The cages were attached at 61 m intervals along 
a groundline with anchors and buoy lines at each end 
of the groundline separated by 61 m from the nearest 
cage (Wakefield 4 ). The groundline was set off the stem 
of the tow vessel (RV Pelican) under tension, from an- 
chor to anchor, so that the line was straight and all 
camera cages had the same general geographic orien- 
tation. Each of the cages had an unlit, monochrome 
stereo camera system, known as the Modular Optical 
Underwater Survey System, equipped with 2 Prosilica 
GT 1920 cameras with 2.82-MP resolution (Allied Vi- 
sion Technologies, Stadtroda, Germany) that were rig- 
idly attached to an aluminum frame with a baseline 
separation of approximately 75 cm. Once filming was 
initiated, synchronized stereo photos were taken at 5 
Hz continuously for up to 8 h. In addition, each cage 
contained a dual-frequency acoustic imaging system 
(DIDSON; Belcher, 2002), and the 2 end cages also 
contained a long baseline acoustic beacon to obtain in- 
formation on cage position and to aid navigation of the 
vehicles to within visual range of the camera cages. 
The towed camera vehicle used in the experiment 
collected both video and still optical imagery with for- 
ward- and side-mounted cameras that were angled 
downward approximately 45° from level and illumi- 
nated with 4 continuous 85-W LED lights. The camera 
vehicle frame measured 1.2 m wide by 0.7 m tall by 
1.7 m long and was designed to be towed at approxi- 
mately 1.8 m/s with a target off-bottom distance of 2.5 
m (Fig. 1; Lembke et ah, 2013). The towing cable was 
standard 10.8-mm hydrographic wire that passed over 
an A-frame along the centerline of the vessel. At the 
depths in the experimental area and at a towing speed 
of 1. 5-2.0 m/s, the camera vehicle trailed behind the 
vessel by -40-50 m. 
At each sampling site, the observation cages were 
deployed, then the positions of the end cages were es- 
timated by using a long-baseline system after circling 
the entire array with the vessel. Approximately 1-2 h 
after cage deployment, depending on the time required 
to obtain quality positions, the camera vehicle was 
launched and towed near the cages with a target hori- 
zontal distance of 3-5 m at closest approach to each 
camera. Towing always occurred in the same direction 
when the camera vehicle passed by the cages, and rep- 
licate passes were separated in time by -20-25 min. 
Stereo camera calibration 
Accuracy of estimates of fish position depend on the 
quality and precision of the calibration of the stereo 
camera. Calibration of the Modular Optical Underwa- 
