uracy (Scopélitis, et 
n deep water corals, 
neters, depths which 
Thus, the SeaBED 
) took the images, 
isting random point 
annularis complex, 
a and had a smooth 
overage of the coral 
ich used underwater 
ice, where the video 
face. The video was 
hen used to train the 
oral and algae) and 
The classifier used 
| linear discriminant 
features were inputs 
11 success rate of the 
s, like 3 meters, a 
e living and 75% for 
plored and used in a 
ured the video at a 
ef surface and the 
of coral reefs. They 
ion neural network 
into three benthic 
1. Color and texture 
y were able to obtain 
e not included in the 
for the same set of 
d on satellite images 
ethods — expertise- 
'he expertise-based 
visual analysis and 
1aps were the most 
s for field validation 
used as reference so 
n the validity of the 
d object-based. All 
ss tabulated with the 
ixel-based methods. 
plished using ENVI 
n algorithm, which 
known to be coral 
used the software 
e to objects, which 
le by the user. The 
all agreement of the 
map was better than 
pased classification 
cover may lead to 
es and faster result 
The precision of the 
h high resolution, 
Mgh-accuracy Was 
, 2006). Repetitive 
isured using a high 
resolution multibeam echo sounder (MBES) with a real-time 
long range kinematic (LRK TM) global positioning system. Four 
annual surveys were carried out and in a single survey, where 
seven measurements were acquired, revealed the precision of 
the MBES system, which was £20 cm horizontally and +2 cm 
vertically at a 95% confidence level. In contrast, a lower 
precision was produced when the four annual surveys were 
compared. The horizontal and vertical precision, at a 
confidence level of 95%, was only £30 cm and +8 cm, 
respectively. Still, it was concluded that the full potential of the 
MBES system did not correspond to the precision achieved in 
the study because these measurements could be improved 
through an increase in density of coverage (soundings/square 
meter) by reducing the vessel’s survey speed. 
Integration of high-resolution images with elevation data 
using OBIA. The fusion of RGB imagery and LiDAR with 
OBIA for classification as well as analysis of savanna systems 
were explored in a research (Levick & Rogers, 2006). They 
discovered that high resolution digital color imagery 
complemented with elevation data from Light Detection and 
Ranging (LiDAR) greatly enhanced the landscape’s structural 
description by adding the factor of height. For the traditional 
pixel-based classification techniques (supervised classification, 
unsupervised classification, etc.), the complex system’s 
heterogeneity at different scales was problematic. However, the 
object-based approach using the software eCognition was able 
to produce accurate classifications. Results from the study 
suggest that incorporating the component of height in 
classifying images as well as using OBIA increased the 
accuracy of resulting classification maps. 
With this and with the possibility of acquiring very precise and 
reliable vertical and horizontal measurements from the MBES, 
exploring the possibility of combining bathymetric data 
acquired from the MBES and underwater photos may lead to 
higher accuracies in object-based classification. Though spatial 
resolution may vary between the underwater photos and the 
MBES data, the height information and their variation may still 
be used to enhance the classification. Investigating how to 
accurately georeference the underwater videos and mosaics to 
the bathymetric data must also be undertaken in order to get 
reliable results. 
2. MAIN BODY OF TEXT 
2.1 Study area &conceptual framework 
Puerto Galera is a municipality in the northwestern part of the 
province of Oriental Mindoro, Philippines. Its location is at the 
Isla Verde Passage’s southwestern end. Studies in the early 
1980s, found that the Puerto Galera area has one of the highest 
number in the world in terms of marine species.! The Coral 
Garden is a dive spot that is 10 minutes away by boat from 
Puerto Galera mainland and its maximum depth is around 12 
meters. 
Benthic Cover 
Map with 
Higher 
Underwater 
Video Surveys 
and 
Photographs Accuracy 
   
Figure 1 — Final concept applied for the study 
  
1 Wikipedia «http://en.wikipedia.org/wiki/Puerto Galera» 
From the theory of combining RGB imagery and LIDAR data 
with OBIA (Levick & Rogers, 2006), combining high 
resolution bathymetric data and underwater photos with OBIA 
may produce a benthic cover map with higher classification 
accuracy. In this methodology, the videoed transect will also be 
given a geographic location through the bathymetric data, 
which may be very precise given a good set of echo sounding 
and positioning systems, with attachments such as a 
gyrocompass and a motion sensor. Figure 1 illustrates the final 
concept applied due to limitations of the available data. 
2.2 General procedure 
E ^ | Georeferencing | 
Corrections 
  
  
Data Gathering 
  
  
  
    
TTT 
Accuracy Classification Manual 
Assessment Methods Delineation 
  
  
Comparison & 
Conclusion 
  
  
  
Figure 2 - General flow of procedure done in the research 
Figure 2 illustrates the general flow of the procedure done in 
this research and will be discussed in more detail in the 
subsections to follow. 
Data acquisition. Two scuba divers laid a 50-m tape along an 
area with various types of benthic cover and took a video at a 
distance of 3 meters from the reef surface. The recorded video, 
which was taken by swimming along the transect at a constant 
speed, was taken at around 10:00 — 12:00, which was within the 
time for best lighting conditions that is between 08:30 — 15:30 
(Hill & Wilkinson, 2004). The instrument used to take photos 
and videos was a Canon S95 placed in an underwater casing, 
with the setting of the camera at video mode, automatic focus 
and the resolution setting was 1280x720 at 24 frames per 
second. The multibeam echo sounder ES3 was used to gather 
bathymetric data over a large area, which covered the 
transected line. It was setup on a boat, which moved at a slow 
speed in order to obtain data with high density. Two dual 
frequency antennas for the Trimble ® SPS461 DGPS 
(differential geographic positioning system — GPS) Beacon 
Receiver were placed on each end of the boat to track the 
movement of the vessel as well as to give geographic location 
to the bathymetric data measured by the MBES. It was used in 
the mode real-time kinematic (RTK) satellite navigation, thus 
giving it a centimeter level accuracy. A high resolution 
QuickBird image was also used during the survey in order to 
double check the geographic location being recorded. Lastly, a 
hand-held GPS was used to record the location of the drop off 
point of the divers. 
Pre-processing of handheld GPS and MBES data. The hand- 
held GPS data was re-projected using the re-project feature 
command of ArcGIS 9.3 in order to convert the point to UTM 
Zone SIN. A tide predictor program | (WXtide47 
[www.wxtide32.com]) was used to produce a tide chart that 
was used to compute tide correction values. These values were 
used to adjust and correct the data acquired by the MBES. An 
approximation of the length and width of the reference coral