338_2016_1522_MOESM1_ESM

Electronic Supplementary Material
Camera used
As camera, we used a GoPro HERO4 Black with a modified lens. The modified lens
substitutes the original GoPro one, and has the following characteristics:

Focal length 5.4mm

Horizontal opening angle 60°

Aperture range F 2.5
Agisoft Photoscan workflow
Workflow, parameters and algorithms used in Agisoft Photoscan are shown in Table S1. The
parameters used in the automatic classification on the dense point cloud in Agisoft Photoscan
(function “Classify Ground Point”) are shown at the bottom of the table. In the first step of the
classification, the dense cloud is divided into cells of a dimension of the parameter “cell size”.
In each cell, the lowest point is detected and a first approximation of the bottom is given by a
triangulation of these points. In the second step a new point is added to the bottom point class
providing that it satisfies two conditions: it lies within a certain distance from the bottom
model (max. distance parameter) and that the angle between the bottom model and the line
connecting the new point with the bottom is less than the “max. angle” parameter. This second
step is repeated while there are still points to be checked. The results of the classification are
shown in Fig. 3f.
1
Table S1 Workflow and parameters used in Agisoft Photoscan to process drone imagery.
First step: Align photo
Accuracy
High
Pair preselection
Disabled
Key point limit
40000
Build mesh (preliminary step to insert GCPs)
Surface type
Arbitrary
Face count
Medium (30000)
Source data
Sparse cloud
Interpolation
Enabled
Point classes
All
Second step: Locate and place GCPs in the scene and
import GCPs coordinates
Measurement accuracy (for GCPs)
Camera accuracy
10
(m)
Marker accuracy
0.05
(m)
Scale bar accuracy
0.001
(m)
Projection accuracy 0.1
(pix)
Tie point accuracy
4
(pix)
Third step: Build dense cloud
Quality
High
Depth filtering
Aggressive
Fourth step: Build mesh
Surface type
Arbitrary
Face count
Medium (136012)
Source data
Dense cloud
Interpolation
Enabled
Point classes
All
Generate Orthophoto – Export Orthophoto
Generate 3D Model – Export DEM
Classification of ground points
(geometrical filtering)
Parameter
Value
Max angle (deg)
0.5
Max distance (m)
0.1
Cell size (m)
1
2
LiDAR data
The LiDAR point cloud was derived from a topobathymetric campaign conducted between 10
and 26 June 2015 by the joint collaboration of Fugro LADS corporation, the French
Polynesian Service de l’urbanisme (SAU), the French Service hydrographique et
océanographique de la marine (SHOM) and the MooreaIDEA consortium funded by the US
National Science Foundation Long Term Ecological Research Program. Moorea’s coastal
fringe was surveyed by the RIEGL VQ-820-G hydrographic airborne laser scanner operating
at 532 nm and 251 kHz with a nominal swath width of 375 m, which leads to a sounding
density of nominally four points m–2. Data were post-processed relative to the International
Terrestrial reference Frame 2008 and delivered to the projection system UTM 6S associated
with the geodetic system RGPF and altimetric system NGPF. The horizontal and vertical
accuracies of 1 m and 0.25 m were computed based on fifty benchmarks.
SfM-derived bathymetry–LiDAR alignment
The SfM-derived bathymetric DEM was aligned to the LiDAR on the basis of the position of
four conspicuous coral heads, shown in Figure S1. The XYZ distances between the points in
Figure S2 are shown in Table S2.
3
Fig. S1 Left panel: the bathymetric DEM obtained from the Agisoft Photoscan workflow.
Right panel: comparison between the DEM in the left panel and the LiDAR dataset (dots, in
tones of grey). The color scale of the DEM is equivalent to that used in Fig. 1f and Fig. 3b–e.
The color scale used for the LiDAR point cloud is a simple grayscale where white represents
deeper areas and black represents shallower ones. The green and orange dots in the right panel
represent the georeferencing points identified respectively in the DEM (green) and in the
LiDAR (orange) for alignment. The results of the alignment process are shown in Table S2.
4
The latitude-longitude differences between the points in the SfM-derived bathymetry obtained
in this study and the LiDAR are shown in Table S2. We used the points in Table S2 to align
the SfM-derived bathymetry to the LiDAR with the ArcMap Georeferencing tool, 1st order
polynomial transformation.
Table S2 Points used to align the SfM-derived bathymetry and the LiDAR cloud point.
Point
ID
1
2
3
4
Longitude
DEM (deg)
Latitude
DEM (deg)
Longitude
LiDAR (deg)
Latitude
LiDAR (deg)
-149.899896
-149.89994
-149.89995
-149.899807
-17.486611
-17.486472
-17.487035
-17.485864
-149.899887
-149.899928
-149.899945
-149.899788
-17.486606
-17.486466
-17.487034
-17.485853
Averages
Delta
X (m)
-1.0
-1.3
-0.6
-2.1
-1.2
Delta
Y (m)
-0.5
-0.6
-0.1
-1.2
-0.6
Delta
distance
(m)
1.1
1.5
0.6
2.4
1.4
5
Agisoft Photoscan report
The flowing information has been extracted from the Agisoft Photoscan processing report.
Figure S2 Camera locations (dots) and image overlap (colors)
Table S3 Details of the flight and on the processing results
Number of images: 306
Camera stations: 306
Flying altitude: 29.4 m
Tie points: 4911
Ground resolution: 7.84 mm/pix
Projections: 86,269
Coverage area: 8.38×103 square meters
Reprojection error: 1.11 pixels
6
Figure S3 Image residuals
Table S3 Details of the flight and the processing results
Type: Frame
Skew: -3.12523
Fx: 3481.7
Cx: 2029.53
Fy:3479.07
Cy: 1520.39
K1: -0.0927356
P1: -0.00105114
K2: 0.142992
P2: -0.000219611
K3: -0.0259784
P3: -2.74474
K4: -0.00238544
P4: 0.744745
7
Table S4 Internal error associated with each Ground Control Point, and total internal errors.
Label
XY error
(m)
Z error (m)
Error (m)
Projections
Error (pix)
GCP_green
GCP_blue
GCP_pink
Sand_end
Sand_center_1
Sand_center_2
Left_sand
Coral_left
Near_boat
GCP_green
TOTAL
0.143012
0.182276
0.0292783
0.652693
0.0799207
0.386935
0.322584
0.941053
0.475707
0.143012
-0.112925
-0.0976561
0.127223
0.0664985
0.0110363
-0.0638596
0.226539
0.333445
0.176596
-0.112925
0.182222
0.206788
0.130548
0.656072
0.0806791
0.392169
0.394183
0.998382
0.507428
0.182222
70
37
41
20
24
21
7
13
5
70
0.696
0.457
0.408
0.011
0.024
0.021
0.088
0.017
0.009
0.696
0.453699
0.163567
0.482283
0.452
Table S5 Internal error associated with each Scale Bar, and total internal errors.
Label
Pink_diagonal
Blue_diagonal
Green_diagonal
TOTAL
Distance (m)
0.594472
0.598206
0.594201
Error (m)
-0.0655278
-0.0617939
-0.0657986
0.0643993
Processing parameters
General
Cameras
Aligned cameras
Markers
Scale bars
Coordinate system
Point Cloud
Points
RMS reprojection error
Max reprojection error
Mean key point size
Effective overlap
Alignment parameters
Accuracy
Pair preselection
Key point limit
Tie point limit
Constrain features by mask
Matching time
Alignment time
Optimization parameters
Parameters
306
306
15
3
WGS 84 (EPSG::4326)
4,911 of 18,067
0.712232 (1.11005 pix)
12.418 (15.0632 pix)
1.63915 pix
30.5845
Highest
Disabled
40,000
1,000
No
6 hours 30 minutes
3 minutes 48 seconds
fx, fy, cx, cy, skew, k1-k4, p1, p2, p3, p4
8
Optimization time
Dense Point Cloud
Points
Reconstruction parameters
Quality
Depth filtering
Processing time
DEM
Size
Coordinate system
Reconstruction parameters
Source data
Interpolation
Orthomosaic
Size
Coordinate system
Channels
Blending mode
Reconstruction parameters
Surface
Enable color correction
8 seconds
55,506,938
High
Aggressive
6 hours 43 minutes
6,275 x 16,621
WGS 84 (EPSG::4326)
Dense cloud
Enabled
8,794 x 26,786
WGS 84 (EPSG::4326)
3, uint8
Mosaic
DEM
No
9