Multi-date correlation of Terrestrial Laser Scanning data for the characterization of landslide kinematics EGU2011-10302-1
Contact: Julien Travelletti Institut de Physique du Globe de Strasbourg, CNRS UMR 7516, University of Strasbourg (EOST), 5 rue René Descartes, 67084 Strasbourg Cedex, France Email:
[email protected]
J. Travelletti (1,2), J.-P. Malet (1), C. Delacourt (3) (1) Institut de Physique du Globe de Strasbourg, CNRS UMR 7516, University of Strasbourg (EOST), 5 rue René Descartes, 67084 Strasbourg Cedex, France (2) GEOPHEN-LETG, CNRS UMR 6554, University of Caen Basse-Normandie, 14032 Caen Cedex, France (3) Université Européenne de Bretagne, Institut Universitaire de la Mer, CNRS UMR 6538, University of Brest, Brest, France
27 May 2010
400
600
600
800
800
v-axis (pixel)
200
400
1000
1000 1200 1400
1400
1600
1600
1800
1800 200
400
600
800
1000 1200 1400 1600 1800 u-axis (pixel)
2000 2200
2400
200
400
600
800
1000 1200 1400 1600 1800 u-axis (pixel)
2000 2200
2400
Fig. 3. Images derived from the gradient calculation on the TLS point clouds at different dates.
A
C
2
Distan ce
0m
(u,v)
k
k v
TLS data ALS data
i Local coordinate system
j
j (XTLS, YTLS,ZTLS) i TLS coordinate system
Fig. 2. Alignment of the stable parts of the different TLS point clouds on the ALS point cloud, with indication on the projective projection applied on the TLS point clouds with the different coordinate systems involved in the procedure.
The result of the correlation corresponds to the displacements Δu and Δv along the u-axis and v-axis. To reconstruct the 3D displacement field, a triplet of coordinates (X, Y, Z) obtained from the first and the second TLS acquisitions is associated to each initial (u,v) and final position (u+Δu , v+Δv) of the displacement vectors in the images. The median of the displacement components (ΔX, ΔY , ΔZ) located in the same mesh (one-meter grid mesh size) is then computed.
0.5
0.004
Dista
nce X (m
)
8 6 4
0 0
0.040
F
C f(x) = 0.99 x - 0. 009 r2 = 0.99
2
5m
0 -0.002
0.030
0.025
0.002
2 0.015
0.010 -0.004 -0.006
2
5m
Distan ce
X (m)
3
)
1.5 1 0.5 0 0
0.8
0.8
0.6 0.5 0.4 0.3 0.2
1 2 3 4 Displacement correlation (m)
0.7 0.6 0.5 0.4 0.3 0.2 0.1
0.2 0.4 0.6 0.8 Displacement correlation (m)
Block 2 Block 3 Block 4 Block 5
0 0
12 Oct. 07 - 19 May 08 19 May 08 - 23 July 08 23 July 08 - 18 Oct. 08 18 Oct. 08 - 12 May 09 12 May 09 - 29 May 09
0.2
0.4
0.6
0.8
Displacement correlation (m)
29 May 09 - 12 July 09 12 July 09 - 24 July 09 24 July 09 - 05 Oct. 09 05 Oct. 09 - 27 May 10
References:
Dista
Dista nce X (m
2
0.9
0.7
f(x) = 0.99 x - 0. 006 r2 = 0.99
3 2.5
4 6 8 10 12 Displacement correlation (m)
Block 1
0
-0.010
4 3.5
0.9
0 0
0.005
-0.008
0
(m) ce Y
)
12
0.1
Elevation Z (m)
u
12
)
(m)
Image perpendicular to the viewing direction
(X, Y, Z)
Dista nce X (m
D
tanc eY
Stable parts
The image correlation principle consists in recognizing identical intensity distribution patterns in a correlation window over two images to determine the displacement of the center of the window by maximizing a normalized cross correlation function (Lewis, 1995).
0.008
Block 1
10
3
1
0.010
0.006
Dis
Each TLS dataset is aligned with a 3D error of +- 4 cm on the stable parts surrounding the landslide. For the georeferencing, an Airborne Laser Scanning (ALS) point cloud was used as a reference (Fig. 2). The georeferencing accuracy presents an average error of +- 1 cm and a standard deviation of 14 cm.
24 July 2009 - 5 Oct 2009
Co-registration and georeferencing
X (m)
B
Elevation Z (m)
Determination of the displacement field by image correlation
1
3
1
1.5
Block 3
Dista nce X (m
B
1
2
Block 4
Block 2
Fig. 5. Tensile fisures (1) and shear fissures (2, 3) observed at the front of the toe (Fig. 4C, E) on a very-high resolution orthophotograph acquired in October 2008 (Niethammer et al. , 2010).
E
Block 5
tan
2400
Displacement ICP (m)
2000 2200
Displacement ICP (m)
1000 1200 1400 1600 1800 u-axis (pixel)
(m)
800
tan ce Y
600
Dis
400
Shear (-)
29 May 2009 200
1200
The monitoring has been realized with a long-range terrestrial laser scan Optech ILRIS-3D (Fig. 1). Ten acquisitions were acquired between October 2007 and May 2010 for the same base station at an average distance of 100 m from the landslide. The point clouds are constituted of 9 to 12 millions of points distributed on a surface of 6’000 m2 (average point density of 230 pt.m2)
200
(m)
2400
ce Y
2000 2200
Dis tan
1000 1200 1400 1600 1800 u-axis (pixel)
Elevation Z (m)
800
Elevation Z (m)
600
(m)
400
Dis tan ce Y
Acquisition of TLS data
1800 200
Surface strain εS (-)
1600
(m)
1600
In 2008 and 2009, the upper part of the toe is characterized by a succession of approximately parallel bands (width of 5 to 10 m) in compression and extension which main orientation is perpendicular to the sliding direction (Fig. 4C, D). The behavior at the front of the toe is very different for the two years. In 2008, the extension behavior creates tensile fissures and important shearing located along the landslide boundary (Fig. 4E). This confirmed by the presence of very persistent tensile and shear fissures observed in the field from veruhigh resolution images (Fig. 5). In 2009, the deformation affecting the front toe is less important (Fig. 4F).
ce Y
1400
tan
1200
1400
Dis
1200
1800
This work presents a simple method to characterize the 3D displacement field of landslide by exploiting all geometrical information contained in 3D point clouds with a normalized cross correlation function. The toe of the Super-Sauze landslide (South French Alps) is chosen as an experimental site (Fig. 1). The landslide extents over a distance of 900 m between an elevation of 1980 m at the crown and 1760 m at the toe with an average width of 135 m and a average slope of 25° (Malet et al., 2002).
800 1000
Elevation Z (m)
800 1000
A Fig. 6. Comparison of the displacements obtained by correlation with the ICP method and the dGPS monitoring. A) Locations of the blocks in the point cloud; B) Comparison with the ICP method; C) Comparison with the dGPS monitoring.
Dis
600
Displacement DGPS (m)
400
600
Displacement DGPS (m)
400
For the period July-October 2008, displacements between 0.5 and 1.5 m are observed (average displacement rate of 0.6-1.7 cm.day-1). The largest displacements are detected at the front of the toe (Fig. 4A). For the period July-October 2009, shorter displacements ranging from 0.1 m at the front of the landslide toe to 0.6 m in the upper part of the toe are observed (average displacement rate of 0.1-0.8 cm.day-1; Fig. 4B).
(m)
Objective
200
v-axis (pixel)
v-axis (pixel)
Scanned area
v-axis (pixel)
A Marie Curie Research & Training Network
Mountain Risks: 2007-2010
Toe, 1760 m
200
The computed displacements are in very good aggreement with the displacements calculated with the Iterative Closest Point algorithm (Besl and McKay, 1992; average error of 2 cm and standard deviation of 2 cm; Fig. 6B) and observed from the GPS monitoring of five blocks at the landslide toe (average error of 4 cm and standard deviation of 3 cm; Fig. 6A, C).
Kinematic analysis
19 May 2008
12 Oct. 2007 Fig. 1. View of the SuperSauze landslide from the TLS base station (laser scan Optech ILRIS-3D). The picture (18 Oct. 2008) corresponds to the plan perpendicular to the viewing direction of the laser scan.
Validation of the computed displacements
Elevation Z (m)
A strain analysis is applied to characterize possible different behaviours in the landslide material. A local least square fitting technique is used to compute the Cauchy’s strain tensor E at each location of the grid (Pan et al., 2009). The two eigenvalues e1 and e2 of E corresponding to the change of length per unit of length in the direction having the maximum and minimum extension (positive for extension) without shearing are computed. The deformation is presented by the surface strain defined as εS = e1 + e2 (positive for extension) and the shear strain defined as = | e1 - e2|.
Dis tanc eY
Stable part
Strain computation
3D displacements (m)
120 m
900 m
50 m
Simplification of the 3D matching problem A projective transformation allows to represent the entire geometrical information in a plan perpendicular to the viewing direction of the scan. The relationship among the 3D position (X, Y, Z) of each point in the ground reference system to its position (u,v) in the plane reference system is given by the collinearity equations (Kraus and Waldhaüsel, 1994) (Fig. 2). A scaling factor is used to adjust the average point density to 1 pt.pixel-1 to preserve the geometrical information. The 2D gradient of the distance between the point clouds and the TLS station is calculated for emphasizing the morphology of the landslide. The generated images are then converted in grey-scale values (Fig. 3).
23 July 2008 - 18 Oct 2008
Scarp, 1980 m
N 44°25’
Elevation Z (m)
6°40’
nce X (m
)
Fig. 4. Results of the displacement and strain calculations for the periods July-October 2008 and July-October 2009. A), B) Calculated 3D displacement field. The shallow slope failure is marked with a dashed line; C), D) Calculated surface strain (positive for extension, negative for compression); E), F) Calculated shear strain. The locations 1, 2 and 3 in C) and E) are related to Fig. 5.
5m
Besl, P., McKay, N, 1992. A method for registration of 3-D shapes. IEEE Trans Pattern Anal Mach Intell, 14, 239-256. Kraus, K., Waldhäusl, P., 1994. Photogrammetry, Fundamentals and Standard processes. vol 1. Hermès (Eds.), Paris. Lewis, J.P. 1995. Fast normalized cross-correlation. Vision Interface, 120-123. Niethammer, U., Rothmund, S., James, M. R., Travelletti, J., Joswig, M. 2010. UAV-based remote Sensing of landslides. International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol. XXXVIII, Part 5 Commission V Symposium, Newcastle upon Tyne, UK. Malet, J.-P., Maquaire, O., Calais, E., 2002. The use of global positioning system techniques for the continuous monitoring of landslides. Geomorphology, 43, 33-54. Pan, B., Asundi, A., Xie, H., Gao, J. 2009. Digital image correlation using iterative least squares and pointwise least squares for displacement field and strain field measurements. Optics and Lasers in Engineering, 47, 865-874.
