Use of a remotely operated vehicle as a



Download 1.82 Mb.
Page3/9
Date conversion08.07.2018
Size1.82 Mb.
1   2   3   4   5   6   7   8   9

4.3 Data collection

A variety of data are collected by the ROV system including accurate position, speed, distance, temperature, date, and time. These data are collected and recorded (WinFrog .DAT files) at 2-second intervals and provide the information necessary to calculate search effort.


WinFrog can also record text notes as events (pressing F10). The text notes are compiled in a log text file (.LOG) including the date, time, depth, and position of the ROV for each event recorded. The most important data events recorded, the counts of abalone on the transect, are generated in this manner by making a text note when an animal is spotted. Abalone shells and the sighting of any other abalone species are also recorded. In addition, any other pertinent information or events during the dive are recorded. For example, a text note of start transect is made once the ROV has reached operating depth and encountered rocky habitat. If traversing sand or other unsuitable abalone habitat a note of “off transect” is made and “on transect” is recorded when rocky habitat is once again encountered.
Other data collected include VHS and DVCAM video footage. The video footage is recorded continuously throughout the transect. The first data role of the video is to record metadata, generated as text on the ROV console. Another role is to record the position of the lasers fixed to the ROV. These lasers are used for measuring distances, field of view and the sizes of animals.

5. Analysis strategies




5.1 Determining transect area

The key information obtained by the ROV are counts of abalone along the transect line at each depth and the transect length and width. By determining the area of transect surveyed and averaging the number of abalone sited, a mean density, with known variation, can be produced for each depth stratum. With this information an extrapolation can be performed, using the total known area of habitat, to estimate the abalone population; hence testing the first null hypothesis (section 1.2).


Determining transect length, width, and quality with accuracy and precision involves a combination of automated and manual post-processing. The ROV trackline, data stream, and video feeds include “off-transect” components which need to be excluded from the calculation of transect area. Transects can go “off-transect” if the ROV loses the reef, needs to move long distances (>100m) between patches of reef, or is moving too fast to be searching effectively. Similarly, when the ROVs pitch changes radically, for instance to fly over an obstacle, the transect is no longer being searched for abalone and this section needs to be discarded from the area calculation. Transects can also be “off-transect” if the ROV does a complete loop and doubles over the previous path. This makes the sub-sample a pseudo-replicate and it must also be discarded from the transect.
Determining when transects are “on” or “off-transect” can be achieved in three ways: GIS analysis of the trackline over the Bathymetry layer, computer notes made on the WinFrog .LOG file by pushing the F10 button, and by post processing of the video. Of these analyses the post processing provides the most precise and accurate data for most off-transect occurrences. This is because bathymetry imaging is of a lower precision than the video footage and is not useful for pitch or speed, while there is the potential for F10 notes in the .LOG file to be forgotten by the operator during very active ROV flights.
Once those parts of the video footage that are “off transect” have been identified and excluded from analysis, the length and width of the transect can be calculated. Even when “on-transect”, the transect width is variable as the ROV dips and rises to maneuver over the substrate. Using reference lasers mounted in a fixed configuration within the video field of view, the width of these transects can be computed during post-processing. This is undertaken using Quantitative Measurement Software (QMS; Green Sky Imaging, 2004). The software can acquire and process video images based on a photo-overlap, fixed-time interval, fixed-distance interval, or randomly sampling. In the case of our abalone transects, photo-overlaps will be used, as total transect area (τ) rather than sub-samples of the transect are needed for the mean density of abalone () to be determined (See section 2.1.6).
The software uses the geometry of the laser references on each overlapping video frame to determine the range to center of the image, center width of the image, area below center and the x-y scale information over the entire field of view (Figure 3). By using this data from the overlapping video images both length and width and hence transect area (τ ) can be calculated (Figure 3).
The measures of the laser path over the transect need to be further adjusted to take into account the roll of the ROV and the pitch of the camera array. Even though this data is captured every two seconds (unlike the continuous collection of laser sightings by the video-photo overlaps), an extrapolating algorithm in the QMS software provides a correction, filling in the gaps between each “ping” of data.

1   2   3   4   5   6   7   8   9


The database is protected by copyright ©dentisty.org 2016
send message

    Main page