Você está na página 1de 11

THE MEASUREMENT OF LANDFILL

SETTLEMENT USING TERRESTRIAL 3D


LASER SCANNER IMAGING

F. OLIVIER*, D. LHOMME**, J.P. GOURC° AND M. HIDRA°°

* Environment, Energy and Waste Research Centre (CREED) 291, avenue


Dreyfous Ducas, 78520 Limay, France (Post-Doctorate LIRIGM)
** ATM3D, Savoie Technolac,House Boat n°7, 73375 Le Bourget du Lac, France
° LIRIGM Laboratory, University of Grenoble, 38041 Grenoble Cedex 9, France
°° ONYX Sanet, Petits Pourcieux – 26300 Chatuzange-le-Goubet, France

SUMMARY: The paper presents terrestrial laser scanning technology and its application as part
of landfill monitoring, with the purpose of better evaluating waste settlements. The move from a
discrete to a continuous representation of surface settlements is indeed expected to improve the
control of sanitary landfills. Laser scanning is introduced with regard to the aspects of field data
acquisition, processing and calibration. The first two experimental campaigns conducted at
Chatuzange landfill (France) are presented focusing at the present stage of the study on three test
zones located in three different cells. Selected areas were subdivided in 2 m wide elementary
grids in order to assess the average altitude of the ground surface as a function of the point cloud
density. In this respect, first results indicate that a density of 25 up to 50 recorded points / m2
(after treatment) is highly desirable. Also, results of the first two campaigns are illustrated by the
production of 3D elevation models and 2D horizontal profiles.

1. INTRODUCTION

Waste is a very compressible material which, when lifted in high embankments, induces major
technical constraints as well as economic concerns. The wastes enclosed in sanitary landfills are
characterized by differences in composition, age, placement and compaction practices resulting
in non-uniform surface movements (differential settlements). With a typical magnitude of
between 10 and 25 %, post-closure settlements represent in fact a serious threat to the integrity
and the performance of the cap cover and the associated structures.
In the framework of a research program conducted over the last 7 years at the University of
Grenoble, special attention has been given to control and predict waste settlements, including the
monitoring of several landfill sites using conventional geodetic techniques as well as various
internal instruments (inclinometers, buried plates / cells, extensometers) [(Olivier, 2003),
ADEME (2005)]. Concerning the conventional geodetic techniques – triangulation and distance
measured with an electronic tachometer (also called total station) or a Global Positioning System
(GPS), despite their good precision, they are often unable to document the evolution of a landfill
with sufficient resolution to detect problems before they cause major environmental damages.

Proceedings Sardinia 2005, Tenth International Waste Management and Landfill Symposium
S. Margherita di Pula, Cagliari, Italy; 3 - 7 October 2005
 2005 by CISA, Environmental Sanitary Engineering Centre, Italy
Sardinia 2005, Tenth International Waste Management and Landfill Symposium

Beside those techniques, ground-based 3D laser scanning technology represents an alternative


procedure for the measurement of surface displacements. Conceived about 10 years ago, laser
scanning is having growing development with applications in a variety of technical fields
including the monitoring of landslides and cliffs [Bitelli et al. (1999), Collins and Sitar (2005)],
tunnel construction and maintenance, archaeological survey, rehabilitation of historical
monuments, etc. Yet, apart from a specific landfill mining project (Göttingen, Germany)
reported by Kern (2005), this technology has not been assessed yet in the case of real-scale
landfills. In this context, an evaluation of the ground-based laser scanning is being initiated today
as part of a large site investigation conducted at Chatuzange-le-Goubet landfill (France). After
presenting an overview of the technique (acquisition and treatment), we will describe its
application on site and discuss the first results derived from the on-going field campaign.

2. PRESENTATION OF THE TERRESTRIAL LASER IMAGING TECHNIQUE

2.1 General description of the laser scanning instrument


A 3D laser scanner is essentially a range finder that calculates the distance to an object point
along a known trajectory in 3D space (Figure 1a). It consists of an emitting diode that produces a
laser source at a very specific frequency. Pulses are emitted at an extremely rapid rate while a
rotating mirror in the head of the scanner reflects the light along different trajectories in a
vertical plane. Simultaneously, the entire head of the scanner rotates in the horizontal plane
thereby collecting data continuously. For each laser impulse, the scanner collects:
• two angles (θ, ϕ) that give the exact direction of the scanned object ;
• the ‘flight time’ of the impulse to reach a surface and return to the scanner (equivalent to a
distance ρ) and the intensity of the received echo impulse.
The result is the collection of hundred of thousands of points, together called a “point cloud”.
The position of each data point is recorded according to its polar coordinates (ρ, θ, ϕ) and
automatically converted into local Cartesian coordinates (x, y, z). In addition, technology has
advanced to the point that the colour and reflective intensity of the points can also be collected
by determining the wavelengths of laser light not absorbed by the scanned object.

Figure 1. (a) Scanning instrument in operation (b) Cylindrical reflector (diameter φ 50 mm).
Sardinia 2005, Tenth International Waste Management and Landfill Symposium

By comparison to total stations that allow the recording of one physical point every few seconds,
ground-based laser scanning makes possible the measurement of up to 8 000 points per second.
The point cloud that is derived is characterized by a very high spatial density (or resolution) and
a precision typically ranging between 5 and 30 mm – to be compared to the uncertainty of up to
100 mm reported by Carter et al. (1999) for airborne laser scanning.

2.2 Methodology of field data acquisition


Considering a given measurement area, one of the limitations of the 3D laser scanning technique
is that the data collection is only available for surfaces that are within the line of sight of the laser
unit itself. In particular, the scanner will not extract data located behind objects (“shadow
zones”). To avoid this situation, the scanner is moved some distance away from its initial
location to a point where the surface in the shadow zone can be detected. Using multiple
locations provides a convenient way to limit the number of shadow zones while also increasing
the resolution of the data collected.
To capture a large study zone, multiple scans must be conducted over the whole area. They
are then registered together by means of reflectors (Figure 1b) placed strategically at several
locations within the landfill, serving as tie-points for merging the data and generate a single 3D
model.

2.3 Data processing


The first step of the data analysis consists of producing a Digital Surface Model (DSM) from
recorded point clouds. After merging all datasets into a single point cloud model with a common
coordinate system, the local coordinate system is transformed into a geodetic system by using
ground control points measured by traditional levelling. Once the coordinates are established, the
second step of the data analysis is to obtain a Digital Terrain Model (DTM) – also called Digital
Elevation Model (DEM) – that reproduces the natural surface of the ground, which implies
filtering of the vegetation and surface structures and data reduction.

3. LASER SCANNING EXPERIMENTATION AT CHATUZANGE LANDFILL

3.1 Presentation of the landfill


The operation of Chatuzange landfill (Onyx) started in December 1992 and is due to continue
approximately until 2021. During the first 13 years of operation, it has received annually
between 110 and 160 000 tons of waste stored in 5 cells as follows:
• Cells A, B and C : totally operated and covered (Table 1) ;
• Cell D and E: partially operated and temporary covered ;
• Cell F: presently in operation.
With regard to cells A, B and C, waste heights vary to a large extent as a result of the particular
shape of the cells (side-stepping). Settlements have been monitored for the last 8 years by means
of surface plates installed provisory immediately after the lower operation stage, then removed
before the start of the upper operation stage and later re-installed at the same horizontal position
above the complete column (just below the final cap cover) (Table 1) (Olivier et al., 2003). In
addition, internal monitoring instruments (buried plates associated with floating rods) were also
installed in Cell B (although not discussed here) (Olivier, 2003). Both surface and internal plates
were monitored by traditional levelling using a TopCon GTS801A total station.
Sardinia 2005, Tenth International Waste Management and Landfill Symposium

The site, which is a pilot landfill satisfying to the ISO 14001 standard, is expected to become an
important “field research laboratory” within the framework of the Environment Programme of
the Rhône Alpes region.

Table 1 – Characteristics of Chatuzange landfill cells A, B and C.


Cell A B C
¾ Lower operation stage
- Start of operation 12/1992 07/1996 02/1998
- End of operation 06/1996 01/1998 01/2000
- Initial waste height 20 - 29 m 5 – 29 m 6 - 36 m
- Temporary cap cover thickness (hc1) 1.75 m 0.75 m 1.10 m
- Duration of the settlement monitoring campaign 35 months 37 months 63 months
¾ Upper operation stage
- Start of operation 12/1999 04/2001 2006 ?
- End of operation 04/2000 09/2001 -
- Additional waste height 4 - 4.50 m 4-8m -
- Final cap cover thickness (hc2) 1.30 m 1m -
- Duration of the settlement monitoring campaign 59 months to date 49 months to date -

Test zone 3
(95 m * 0.5 m)
Cell A

Cell F
Cell B

Cell E

Cell D

Test zone 1 Cell C


(20 m * 20 m) Test zone 2
(30 m * 30 m )

Figure 2. Plan views of Chatuzange landfill according to (a) the Digital Surface Model (DSM)
including surface structures (highlighted) and vegetation (b) the Digital Terrain Model
(DTM) with local x, y coordinates (in metres).
Sardinia 2005, Tenth International Waste Management and Landfill Symposium

Figure 3. Laser scanning at Chatuzange landfill (a) sensor embarked on top of a car (Campaign
1: Nov. 2004) (b) sensor mounted on a tripod (Campaign 3: March 2005).

3.2 Field monitoring at Chatuzange landfill


In complement to traditional field survey, three laser scanning campaigns have been carried out
to date (i.e. Campaign 1 on 20-21 November 2004, campaign 2 on 06-07 January 2005 and
campaign 3 on 28-29 April 2005). However, only the results of the first two campaigns are
presented in this paper.
The survey instrument tested is a ground-based Riegl LMS-Z420i laser scanner (of operating
range up to 800 m) associated to a calibrated Nikon D100 6 Mega Pixels digital camera mounted
on top of the instrument (Figure 1a and Figure 3). A standard notebook and the software package
RiSCAN PRO™ enable the user to acquire the 3D data and provide a variety of registration, post
processing and export functions.
The three campaigns consisted of between 27 and 30 “stations” (or locations) of acquisition
(Table 2) distributed in an homogeneous way over the site to cover the whole area (Figure 4a:
stations of acquisition are surrounded by a white halo representing a zone of very high point
density).
Each station was referenced from a network of 6 to 15 cylindrical reflectors (Figure 1b),
surveyed and further readjusted according to the Lambert III coordinate system (with an origin
placed 44.1° North and 2.33723° East). The local coordinates (x, y, z) used hereafter were
translated from the Lambert III geo-referencing system according to the following conversion:
• x = xLambert - 819 400 (metres)
• y = yLambert - 306 600 (metres)
• z = zLambert - 200 (metres)
According to the number of reflectors required for geo-referencing, the distribution and the
distance of these targets as well as the average quadratic error characterizing their mathematical
treatment (between 5 and 18 mm for targets located at a distance of up to 250 m), the range of
each station was restricted to a distance of 150 m (i.e. every point located out of range being
excluded from the modelling).
Sardinia 2005, Tenth International Waste Management and Landfill Symposium

Table 2 – Main parameters associated to the first two laser scanning campaigns.
Parameter 20/11/2004 06/01/2005
- Number of stations 27 30
- Number of points / station 2.88 millions 2.88 millions
- Average quadratic error (Emq2) 5 mm ≤ σ ≤ 15 mm 5 mm ≤ σ ≤ 18 mm
- Number of readjustment points / station between 6 and 12 between 6 and 15
- Number of points considered 23 millions 24 millions
- Points remaining after filtering and sampling @ 3 cm 8.5 millions 11 millions
- Points remaining after filtering and sampling @ 8 cm 3.3 millions 3.3 millions

Stations of acquisition

Figure 4. Point cloud density (number of points / 4 m2) associated to the laser scanning campaign
1 (a) Initial point density (b) Final point density after filtering and sampling.

For each station, a panorama scan was executed, characterised by a field of view 360 ° x 80°, an
angular resolution of 0.1° and an operation time of around 4 minutes for over 2 millions
measurement points. It was followed by the fine scanning of the reflectors and series of photos
covering the corresponding field of view.

3.3 Treatment of Chatuzange field data


The treatment of the point clouds derived from the two campaigns consisted of the following:
• Stage 1 (pre-filtering of the data): suppression of low isolated points and clusters of low
points (parasitic multi-routes, parasitic reflections).
• Stage 2 (filtering of the ground): the bare earth points were found by using an algorithm
that makes comparisons between the elevations of adjacent laser points and eliminates
points that have elevation differences greater than set criteria (max. difference in height
between two contiguous points: 3 cm).
• Stage 3 (sampling of the data): the initial dataset consisted of randomly spaced points. In
order to reduce the size of the point clouds, the data has been processed into a regularly
Sardinia 2005, Tenth International Waste Management and Landfill Symposium

spaced grid at a desired grid-node spacing of 8 cm (i.e. redundant points being removed).
The dataset has been homogenized in terms of its density according to the roughness of the
relief (plane zones being less dense than disturbed zones). Eventually, a point cloud of
approximately 3.3 millions points was obtained for each campaign (Table 2) representing
an average of 25 points / m² (i.e. that is, referring to Figure 4b, around 100 points / 4 m2).
Following filtering and sampling of the data, a Digital Terrain Model (DTM) was generated with
a grid size of 20 cm (by triangulation of Delaunay) for fully covered cells A, B and C and with a
grid size of 50 cm (by kriging) for other cells.

4. INITIAL RESULTS FROM THE SCANNER LASER EXPERIMENTATION

4.1 Calibration of the laser scanner instrument


With the purpose to validate the vertical accuracy of the laser instrument for a single point
determination (without any account of the effect of the terrain ruggedness and the cover grass),
the traditional survey of 10 reference points located on Cell B (physically represented by the top
of a metallic rod) was conducted in parallel to the laser scanning campaign using a TopCon
GTS801A total station. Elevations (z) were evaluated using both techniques as detailed in Table
3.
Vertical variations zi − zi' were found to range between 2 and 25 mm, equivalent to an average
value of 12 mm. The level of vertical accuracy derived for a single point determination appears
close to the value suggested by Kern (2005). In theory, the precision of single scan data could be
increased to 5 mm by averaging the results of a multiple scanning process (called scan sequence)
(Riegl, 2005). Such values do not reflect however the typical accuracy of landfill measurements.
Indeed, the ground precision is expected to be substantially poorer as a result of:
• the vertical deviation induced by the geo-referencing process (in relation with the
“polygonal leveling” of field reflectors from an absolute reference point located in a non-
settling area) ;
• the effect of the terrain ruggedness and the cover grass in interaction with the filtrating
process: to which extent does the grass density / height affects the measurements ?

Table 3 - Comparison of reference point elevations derived from traditional surveying and laser
scanning (Campaign 2 ; cell B)
Reference point Distance from 3D laser scanner Total station z i − z i'
N° Type scanner (m) x (m) y (m) z i (m) z i' (m) (m)
1 PE14 23.65 279.02 326.20 256.215 256.190 0.025
2 PE13 19.83 289.99 328.65 254.334 254.318 0.016
3 PS13 23.21 299.14 325.05 253.741 253.730 0.011
4 PE10 9.36 296.77 353.92 252.815 252.806 0.009
5 PS5 28.85 294.03 375.38 250.313 250.304 0.009
6 PT5 30.45 294.74 376.91 251.028 251.014 0.014
7 PE1 56.14 291.09 402.85 249.736 249.764 0.028
8 PE7 46.19 253.71 374.24 253.211 253.215 0.004
9 PE11 14.56 277.17 351.84 254.549 254.545 0.004
10 PE6 33.96 274.22 376.35 252.520 252.518 0.002
Sardinia 2005, Tenth International Waste Management and Landfill Symposium

4.2 Requirements related to the level of discretisation


Before relying on the altitudes derived from the laser scanner, it is important to evaluate the
effect of point cloud sampling on the data accuracy. With this aim in view, a delimited zone of
the landfill of dimension 20 * 20 m (Test zone 1 located on cell C: see Figure 2) was considered,
further discretised into 100 elementary grids of 2 * 2 m (4 m2) (Figure 5a).
Given a certain elementary grid (after filtering of the data), let us consider the random
selection of p points (1 ≤ p ≤ n) amongst the n points located inside the grid. The average altitude
relatively to the n (respectively p) points is given by:
1 n
zn = ∑ zi
n i =1
(1)

1 p
zp = ∑ zi
p i =1
(2)

Looking at Figure 5b, the range of variation of n appears comprised between 64 and 464 (points /
4m2). Yet, only 9 elementary grids (in dash) present a value of n < 100. Corresponding average
altitudes are also given for every grid. Now, it is proposed to determine a threshold for p
(number of points) under which z p can not be considered as significant. One possible criteria
would be to find out the number p such that ψ = z n − z p ≤ ε where ε represents the level of
fluctuation considered as acceptable for the average altitude z p . In order to determine this
threshold, ψ has been plotted as a function of p (varying between 5 and 400) for 10 randomly
selected grids (Figure 6). For p ≥150 (corresponding to a density of 37.5 points /m2), ψ does not
exceed 10 mm. For p ≥ 200, ψ remains under 5 mm further indicating the stabilisation of z p .

a b c d e f g h i j

a1 j1 1
54.489 54.495 54.501 54.498 54.515 54.546 54.592 54.620 54.642 54.648

436 417 447 446 400 388 352 360 346 303

54.279 54.320 54.353 54.382 54.406 54.454 54.482 54.503 54.521 54.557
2
429 396 391 410 398 375 324 382 320 292

54.181 54.223 54.262 54.278 54.291 54.322 54.345 54.364 54.409 54.425
3
326 310 367 390 450 464 447 439 392 349

54.098 54.114 54.118 54.119 54.125 54.153 54.156 54.181 54.209 54.216
4
317 337 360 419 429 436 438 389 363 357

53.999 53.999 53.988 53.966 53.951 53.957 53.959 53.991 54.036 54.070
5
295 327 343 352 364 317 355 349 328 330

53.878 53.864 53.843 53.820 53.810 53.800 53.815 53.841 53.882 53.930
6
273 269 271 217 207 213 217 257 276 281
a10 j10
53.773 53.763 53.743 53.717 53.685 53.716 53.749 53.762 53.807 53.861
7
216 210 190 201 87 76 99 187 192 220

53.732 53.714 53.678 53.658 53.654 53.657 53.683 53.697 53.788 53.880
8
181 160 216 204 198 173 138 73 120 120

53.711 53.715 53.662 53.629 53.630 53.615 53.631 53.675 53.745 53.836
9
107 85 133 147 160 173 150 151 119 85

53.671 53.696 53.635 53.655 53.684 53.573 53.536 53.566 53.609 53.660
10
150 102 81 228 288 157 84 64 76 109

Figure 5. Test zone 1 (20 m * 20 m) (a) Discretisation in 2 m * 2 m elementary grids (b) Average
altitude ( z n ) and number of points (n) associated to each elementary grid (after
filtrating) (Nov. 2004).
Sardinia 2005, Tenth International Waste Management and Landfill Symposium

Average difference in altitude (ψ) (ψ) envelope curve

ψ = 10 mm

ψ = 5 mm

Figure 6. ψ = z n − z p versus the number of points (p) for 10 randomly-selected elementary grids.

The above results were counter-checked considering in a similar way the other 90 elementary
grids composing Test zone 1. Similar trends were observed, confirming the necessity to guaranty
a minimum point density (after filtering) according to the overall accuracy expected.

4.3 Application prospects for the scanner laser technique


Due to the short period of time ( 6 weeks) separating Campaigns 1 and 2 and the late analysis of
Campaign 3, movements observed on cells A, B and C are not significant enough to be presented
hereafter. At this stage, only the future application prospects of the laser scanner technique are
therefore presented focusing on two aspects as follows:

¾ 3 D application: description of altitude variations within a cell


54.3

54.2
55 54.1

54
54
z (m)

53.9

53.8
53
140 53.7

53.6
130
53.5
160
53.4
120 150
53.3
140
y (m)
110 130 x (m)

Figure 7. Representation of the average altitude ( zn ) of Test zone 2 (30 * 30 m) (130 m ≤ x ≤ 160
m ; 110 m ≤ y ≤ 140 m) discretised into 2 m * 2 m elementary grids.
Sardinia 2005, Tenth International Waste Management and Landfill Symposium

51
Slope separating (x, y) = (200, 460)
50
cells A & B
49

48

Altitude (m)
Cell B Cell A
47

46
y = - 1.6 x + 780
45

44
(x, y) = (250, 380)
43
0 10 20 30 40 50 60 70 80 90 100
Distance (m)

Figure 8. (a) View of the slope separating cells A and B in the background (b) Horizontal profile
along the slope separating cells A and B.

Using a software dedicated to graphical applications, various types of contour / colour maps
representative of the terrain relief can be produced such as the one presented on Figure 7. Given
for instance a study zone subject to intensive collapse (Test zone 2 ; see Figure 2b), the
comparison of altitudes between two dates would precise the magnitude of settlement while
describing spatially the extent of differential settlement taking place. To that respect, illustrating
the possible effect of vertical wells – dedicated to the collection of biogas or the recirculation of
leachate (in the case of bioreactors) – on settlement can also be very interesting.

¾ 2D application: description of altitude variations at the interface of adjacent cells


Considering the boundary of two adjacent cells (cells A and B for instance), their vertical cross
section can be observed (Test zone 3 ; see Figure 2b) along a selected horizontal axis, hence
providing detailed information on slope gradient variations (Figure 8). Now, the comparison of
identical profiles at successive dates can document the evolving waste movements at the
interface between the cells. Similarly, the presence of superficial slip features along lateral slopes
may be highlighted by selected 2D profiles.

5. CONCLUSIONS AND PROSPECTS

Ground-based 3D laser scanning technology represents a new method of surface modelling for
landfill applications. Where high resolution is required, point density and ease of implementation
are good assets. By comparison to conventional surveying, the application of the scanner
technique in landfills is expected to improve the assessment of cell capacity (before and during
filling operation) as well as the surveillance of covered cells during the post-capping period.
With regard to the surface structures (biogas drainage vents, rainwater drains, earth slopes,
dikes), the move from a discrete to a continuous representation of surface settlements (with
millions of points captured) will surely result in a better understanding of localised hydro-bio-
physical and mechanical interactions between waste and structures. In this regard, the laser
scanning technique offers a promising opportunity that should result in improved management of
landfills.
Focusing on the present research, further survey campaigns are foreseen that should result in a
more in-depth analysis of settlements based on the scanned data. The effect of the terrain
ruggedness and the cover grass on the data accuracy will be particularly controlled. A
geostatistical analysis is also expected in order to estimate the spatial distribution of settlements
and hence better depict the heterogenous nature of the waste material. Lastly, outputs from
Sardinia 2005, Tenth International Waste Management and Landfill Symposium

settlement variation maps and profiles will be discussed in the light of the conventional data
collected for the last 8 years and further analysed using the Incremental Settlement Prediction
Model (ISPM) [(Olivier, 2003), (ADEME, 2005)].

ACKNOWLEDGMENTS

The authors wish to thank Mr Abdessahel Zerouali (student at the University of Grenoble) and
Loïc Galisson (ATM3D) for their valuable help to the analytical work.

REFERENCES

ADEME (2005) Guide méthodologique pour le suivi des tassements des Centres de Stockage de
Classe II. <www.ademe.fr/htdocs/publications/publipdf/tassements.htm>
Bitelli, G., Dubbini, M. and Zanutta, A. (2004) Terrestrial laser scanning and digital
photogrametry techniques to monitor landslide bodies. Proc. ISPRS Congress, International
Society for Photogrammetry and Remote Sensing, Istanbul, Turkey.
Carter, W., Shrestha R., Bloomquist, D., Lee, M., Finer, P., Sartori, M. and Arnold, D. (1999)
Instant snapshot of landfills using airborne laser swath mapping. Internal report, University of
Florida, Department of Civil Engineering, Geomatics Program.
Collins, B.D. and Sitar, N. (2005) Monitoring of coastal bluff stability using resolution 3D laser
scanning. Proc. GeoFrontiers 2005, Austin, USA, CD-Rom.
Hashash, Y.M.A., Oliveira Filho, J.N., Su, Y.Y. and Liu, L.Y. (2005) 3D laser scanning for
tracking supported excavation construction. Proc. GeoFrontiers 2005, Austin, USA, CD-
Rom.
Kern, F. (2005) Precise determination of volume with terrestrial 3D laser scanner. Proc.
GeoFrontiers 2005, Austin, USA, CD-Rom.
Olivier, F. (2003) Tassement des déchets en CSD de classe II : du site au modèle. PhD thesis
dissertation, University of Grenoble, 325p.
<www-lirigm.ujf-grenoble.fr/Laboratoire/Personnel/folivier.html>
Riegl (2005) Laser measurement systems. <www.riegl.com>.

Você também pode gostar