70 
Fishery Bulletin 99(1 ) 
Ideal edge 
Image profile 
Image profile 
Actual edge a> 
Typical location 
manually 
selected 
Image profile and 
sampled pixels 
High contrast image High contrast image 
Figure 8 
The effects of different contrast in images for location of the edges of objects. 
divers, and diving equipment. Shore-based diving would 
be extremely difficult. 
Before a stereo-video system is used to collect and mea- 
sure field data, operators should practice taking measure- 
ments in the laboratory with a set of test objects of known 
size, so that they may learn to distinguish the edge of the 
object accurately within a range of water clarity. 
Conclusions 
Many of the problems outlined in our study can be over- 
come in the near future. The development of digital video 
cameras and frame grabbing boards have improved image 
quality and the ease of image acquisition and synchroni- 
zation. These developments will also facilitate the analy- 
sis of streams of video images rather than single images, 
thereby greatly increasing the speed of image processing. 
In the future, the combination of stereo-videography with 
neural networks and fuzzy logic could facilitate automated 
pattern recognition, classification, and measurement of 
reef fish from video images. 
Worldwide, temperate and tropical reefs are being threat- 
ened by anthropogenic disturbances. There is a need to 
describe the ecological structure and function of reef-fish 
assemblages and to monitor the effects of disturbance on 
these populations. This type of research has traditionally 
been carried out by government agencies and academic in- 
stitutions. However, limitations in funding and resources 
are forcing these agencies to use supplementary sources 
of data. Consequently, the use of volunteers in monitoring 
programs is increasing (Halusky et ah, 1994; Mumby et 
ah, 1995; Darwall and Dulvy, 1996). Hunter and Maragos 
(1992) suggested that new technology in computing and 
underwater video systems may allow recreational and vol- 
unteer SCUBA divers to assist with surveys of coral reefs 
without compromising the data quality. We have demon- 
strated that volunteers can in fact carry out surveys of the 
length of reef fish using a stereo-video without compromis- 
ing data quality. Although the differences in the accuracy 
of stereo-video measurements made by an experienced op- 
erator and volunteers is statistically significant, these are 
negligible in comparison to errors in visual estimates by 
divers. Our study specifically addresses length estimates 
and does not address estimation of abundance. However, 
if someone is estimating all the lengths of a fish observed 
within a sample unit they are also effectively recording rel- 
ative abundance. The accuracy and precision of abundance 
estimates is an important issue. Factors that need address- 
ing include the accuracy of estimates made by a diver at a 
distance from the fish and the behavioral differences of fish 
at various spatial and temporal scales. These are complex 
issues that are beyond the scope of the present study. 
Quantitative sampling of reef-fish length frequency or 
biomass for monitoring programs requires reliable identi- 
fication skills and the ability to make precise and accurate 
estimates of reef-fish length. It is suggested that volunteer 
SCUBA divers could be trained to use a stereo-video sys- 
tem both in the water and in the laboratory for making 
measurements. Volunteers, under the guidance of profes- 
sional scientists, could assist with monitoring programs of 
reef-fish length frequency and abundance without having 
any effect on the quality of data recorded. The data col- 
lected and analyzed from stereo-video images are proven 
to be considerably more precise and accurate than visual 
estimates undertaken by experienced scientific divers. 
