Harvey et al.: Comparison of estimates of reef-fish lengths made by divers and a stereo-video system 
69 
of statistical power (Harvey et al., 2000). With 
minimal training, volunteers can assist with the 
analysis of images. In addition, a remote stereo- 
video system can be used to record length-fre- 
quency data without harm to the fish from far 
greater depths and over longer periods of time 
than is possible by employing SCUBA divers. 
These advantages will likely have applications 
in fisheries management and deep sea biological 
surveys. 
Unlike a diver who often has to make an im- 
mediate decision on the identity and length of a 
fish, a stereo-video system observer can review 
the images later and repeatedly. Multiple imag- 
es of the same fish are recorded, enabling the se- 
lection of paired images with the best angle of 
orientation to the cameras. Moreover, where im- 
ages of fish are at an acute angle to the cam- 
eras, the system is still able to make accurate 
measurements provided that the head and tail 
of the fish can be seen in both images (Harvey 
and Shortis, 1996). In addition, once an image 
is on screen, multiple measurements of one fish 
can be made over a short period of time (example 
10 measurements in 30 seconds), further reduc- 
ing measurement error. Our results suggest that 
at least five such measurements yield the most 
accurate results. 
Disadvantages of an stereo-video system 
One of the disadvantages of a stereo-video system 
is the financial cost of the equipment: two video 
cameras with underwater housings; a PC com- 
puter; and a suitable frame grabber required to 
convert video sequences to the readable format 
of digital images. Time constraints also need to 
be considered. Although approximately 3 min- 
utes are required to complete a calibration in the 
water, approximately one hour is required in the 
laboratory to capture the 32 images and process them into 
a calibration file (Harvey and Shortis, 1998). The major 
time constraint occurs in the selection and synchroniza- 
tion of the paired video images from the left and right 
videotapes. Once the calibration and synchronization pro- 
cesses are complete, recording of image locations with the 
measurement observation system is reasonably efficient. 
The observer visually locates object features of interest, 
positions a cursor on the feature and clicks the mouse. 
Some physical constraints also need to be considered with 
a design like ours. A bar 1.5 m wide is used to separate the 
video cameras and may become entangled if used in large 
algal beds such as those of Macrocystis pyrifera. In addi- 
tion, the underwater video housings create water resis- 
tance and the system can be difficult to maneuver into 
a strong current. This system is unsuitable for censusing 
cryptic species, because its base separation is too large to 
maneuver into small crevices. The system is best deployed 
off a boat, i.e. a boat large enough to safely carry the stereo- 
rig, a calibration cube (see Harvey and Shortis, 1996), 
0.0 
-0.5 
- 1.0 
-15 
0% 
- 1 % 
■■95 
■■94 
£ 1.0 
3 0.5 
0.0 
■■94 
•■95 
- 2 % 
-3% 
-4% 
-5% 
■■95 
■■94 
6% 
4% 
iS 2% 
0 > 
cr 
■■94 
■■95 
0 % 
Figure 6 
Means and 95% confidence intervals, for each measure of error, for 
1994 and 1995. 
Number of repeat measurements 
Figure 7 
Improvements in the accuracy of stereo-video mea- 
surements due to increasing the number of repeat 
measurements. 
