The main goal of this technique is to produce a 3D model that can be used to derive many parameters (in post-processing) like coral cover (quantitative variable) classified by genus (qualitative variable), rugosity. This technique is applied at exactly the same place every time, allowing a time lapse of the exact same portion of the reef.

## The history

This technique has been used since 2018, following a collaboration with the 100 Island Challenge team lead by Professors Stuart Sandin et Brian Zgliczynski from the Scripps Institution of Oceanography (San Diego, California, USA). Read more about it here.

## The method

The method consists in taking around 2000 pictures of a 10m square (100 m²) portion of the reef. The whole area is positioned following the method used by the 100 Island Challenge team and in the middle of the Photo-quadrat coral transect. The area is covered using an 18mm lens and a 50mm lens.

Illustration of a 3D reconstrction by Visual SFM

## Data processing

• The technique uses the 2000 pictures to generate a 3D model using Agisoft Photoscan in the form of a dense cloud or a mesh model.

• The dense cloud is then fed to a software developped by the Scripps Institution where virtual quadrats can be generated to derive coral cover (and possibly other substrate cover). The software also extracts points along lines on the generated surface that can be processed to obtain rugosity. We extract 20 lines parallel to the photo-quadrat transect and 20 lines perpendicular to the photo-quadrat transect with 1000 points each. We then compute the rugosity as the ratio of a line length to its length project on a mean slope calculated as linear interpolation of the points of the line. Rugosity of the area is then the mean of all lines.

Illustration of transects over 3D model in Viscore

## The products

To compute the rugosity, the following by products are available upon request to Serge PLANES:
- the dense cloud in PLY format (between 50 and 70GB of data)
- a simplified 3D model in OBJ format (between 50 and 100MB)
- an orthomosaic picture in JPG format (around 10MB)
- all the original picture

For a station of 2x20 lines of 1000 points are generate to derive the rugosity
<HR>