Forest vegetation change and affect to biomass of forest vegetation

advertisement
1
Potential of Full waveform airborne laser scanning data for classification
2
of urban areas
3
Tran Thi Huong Gianga*, Nguyen Ba Duyb
4
a
5
Mining and Geology; [email protected]
6
b
7
Hanoi University of Mining and Geology; [email protected]
Cartography Department, Surveying and Mapping Faculty, Hanoi University of
Photogrammetry and Remote Sensing Department, Surveying and Mapping Faculty,
8
9
*Corresponding author:
10
Name: Tran Thi Huong Giang
11
Phone number: +43-6802258679
12
E-mail address: [email protected]
1
13
ABSTRACT
14
In contrast to conventional airborne multi-echo laser scanner systems, full-waveform
15
(FWF) LiDAR (Light Detection and Ranging) systems are able to record the entire emitted
16
and backscattered signal of each laser pulse. Instead of clouds of individual 3D points,
17
FWF devices provide connected 1D profiles of the 3D scene, which contain more detailed
18
and additional information about the structure of the illuminated surfaces. This is an
19
advantage for classification. In Vietnam, airborne LiDAR data primary used for DEM
20
(Digital Elevation Model) generation and almost urban classification researches are based
21
on Remote Sensing data. This paper contributes one potential method to classify urban
22
object based on full-waveform airborne LiDAR data by using additional attributes of full-
23
waveform data such as: sigma0, echo width, echo ratio, amplitude and so on. The output
24
with the high accuracy classification for buildings, trees and water are 93.7%, 92.0% and
25
81.6% respectively. According to this result, the paper demonstrates the potential of FWF
26
data for classification in Vietnam.
27
28
29
30
Keyword: LiDAR, Full-waveform, urban classification
2
31
1. Introduction
32
33
Airborne LiDAR has already proven to be a state-of-the-art technology for
34
high resolution and highly accurate topographic data acquisition with active and direct
35
determination of the earth surface elevation (Vosselman and Maas 2010). Generally, two
36
different generations of receiver units exist: discrete echo recording systems, which are able
37
to record multiple echoes on-line and typically sort up to four echoes per laser shot
38
(Lemmens 2009) and full-waveform (FWF) recording systems capturing the entire time-
39
dependent variation of the received signal power with a defined sampling interval such as
40
1ns (1 nanometer per second)(Wagner et al. 2006, Mallet and Bretar 2009). With signal
41
processing methods, FWF data provide additional information which offers the opportunity
42
to overcome many drawbacks of classical multi-echo LiDAR data on reflecting
43
characteristics of the objects, which are relevant in urban classification.
44
Airborne LiDAR data have been used in various applications in urban
45
environments, particularly aiming at mapping and modeling the city landscape in 3D with
46
its artificial land cover types such as buildings, power lines, bridges, roads. Moreover, as
47
urban environments are active regions with respect to alteration in land cover so urban
48
classification plays an important role in update changed information (Matikainen et al.
49
2010). If FWF data is available, amplitude, echo width, and the integral of the received
50
signal are additional information. Furthermore, a higher number of detected echoes has
51
been reported for FWF data in comparison to discrete return point clouds. These additional
52
attributes were successfully used in classification (Alexander et al. 2010). The
53
classification methods applied reach from simple decision trees to support vector machines
54
(SWM). (Ducic et al. 2006) applied a decision tree based on amplitude, pulse width, and
55
the number of pulses attributes of FWF data in order to distinguish the vegetation points
56
and non-vegetation points. (Rutzinger et al. 2008) used a decision tree based on the
57
homogeneity of echo width to classify points from FWF ALS data to detect all vegetation –
58
trees and shrubs. (Mallet et al. 2008) used SVM to classify four main classes in urban area
59
(e.g. buildings, vegetation, artificial ground, and natural ground).
3
60
In this study, four main classes are derived for the built up areas of
61
Eisenstadt region: buildings, vegetation, water body and ground. They are classified based
62
on decision tree method using OPALS (Pfeifer et al. 2014). The parameters of the
63
classification (threshold values, etc.) are set by expert knowledge or learned from training
64
data. Thus, these values are optimal for the investigated data set.
65
66
2. Study area
67
68
Eisenstadt is a town in the south eastern part of Austria. It characterized by
69
buildings of medium size. This area has landscape is quite similar to urban areas in
70
Vietnam (Figure 1). The center of Eisenstadt was selected for the analzsis located on lat. N
71
47o50’51”, long. E 16o31’5”.
72
73
3. Data used and Methodology
74
75
3.1. Data
76
77
The full-waveform airborne LiDAR data for Eisenstadt area was scanned
78
with a Riegl LMS-Q560 sensor in April 2010. The result point density was about 8
79
points/m2 in the non-overlapping areas, while the laser-beam footprint was not larger than
80
60 cm in diameter. The raw FWF data set were processed by using OPALS software and
81
sensor manufacturer software. The output of this procedure was strip-wise georeferenced
82
point clouds, stored in the OPALS data manager (ODM) format and projected in
83
ETRS89/UTM zone 33N. Each ODM file includes point attributes such as: X-, Y-and Z-
84
coordinate, Intensity, Echo Number, Number of Echoes, Amplitude, Reflectance, Echo
85
Width, etc…
86
Additionally to the LiDAR data, RGB Orthophoto - projected in the same
87
coordinate system - was used for visually interpretation. Furthermore, a Digital Terrain
88
Model (DTM), GEO TIFF format, grid size 1m is also available.
4
89
3.2. Methodology
90
91
First, a number of attributes is computed for each point, using the paradigm
92
of point cloud processing (Otepka et al. 2013) . From these attributes different images are
93
computed (“gridding”) at a pixel size of 1m. Then, a decision tree is applied to classify each
94
pixel into one of the four classes: building, vegetation, ground, and water body. Image
95
algebra (e.g., morphological operations) is used in between to refine the results. The quality
96
of the results is assessed using completeness and correctness measure.
97
98
3.2.1. Attributes for the classification
99
100
Prior to attribute computation in each point, the LiDAR point clouds are
101
checked in order to remove erroneous points which influenced to the accuracy of further
102
processing steps. The relative height of each point above the DTM, nH = z (point) - z
103
(DTM), was computed. All points with nH below -1m and above > 40m are removed. For
104
the Vienna data set the highest buildings are approx. 100m, but also no erroneously high
105
points were found in the data.
106
The value nH defines the attribute nDSM, i.e. normalized surface model
107
(object height). The nDSM represents, as written above, the height of points above the
108
terrain. In the classification it is used to distinguish all the point above the terrain such as
109
buildings and vegetation from the ground points.
110
To distinguish buildings and vegetation points the Echo Ratio (Rutzinger et
111
al. 2008, Höfle et al. 2009) is used. The echo ratio (ER) is a measure for local transparency
112
and roughness and is calculated in the 3D point cloud. The ER is derived for each laser
113
point and is defined as follows:
Echo Ratio ER[%] = n3D / n2D * 100.0
(1)
114
n3D = Number of points within distance measured in 3D (sphere).
115
n2D = Number of points within distance measured in 2D (unbounded
116
cylinder).
117
In building and ground, the ER value reach a high number (approximately
118
100%), but for vegetation and permeable object ER < 100%. ER is created by using
5
119
OpalEchoRatio module, with search radius is 1 m, slope-adaptive mode. For the further
120
analyses the slope-adaptive ER is aggregated in 1 m cells using the mean value within each
121
cell.
122
The attribute Sigma0 is the plane fitting accuracy (std.dev. of residuals) for
123
the orthogonal regression plane in the 3D neighborhood (ten nearest neighbors) of each
124
point. It is measured in meter. Not only the roofs, but also the points on a vertical wall are
125
in flat neighborhoods. Echo Ratio and Sigma0 both represent the dispersion measures.
126
Concerning their value they are inverse to each other (vegetation: low ER, high Sigma0).
127
What is more, Sigma0 is only considering a spherical neighborhood and looks for smooth
128
surfaces, which may also be oriented vertically. The ER, on the other hand, considers
129
(approximately) the measurement direction of the laser rays (vertical cylinder). Those two
130
attributes play an importance role in discriminate trees and buildings. Using OpalsGrid
131
module with moving least square interpolation approach is used to create the Sigma0 image
132
with the grid size of 1 m.
133
The Echo Width (EW) represents the range distribution of all individual
134
scatterers contributing to one echo. The width information of the echo pulse provides
135
information on the surface roughness, the slope of the target (especially for large
136
footprints), or the depth of a volumetric target. Therefore, the echo width is narrow in open
137
terrain areas and increases for echoes backscattered from rough surfaces (e.g. canopy,
138
bushes, and grasses. Terrain points are typically characterized by small echo width and off-
139
terrain points by higher ones. The echo width also increases with increasing width of the
140
emitted pulse. It is measured in nano seconds. OpalsCell module is used to create EW
141
image with the final gird size of 1 m.
142
The local density of echoes can be used for detecting water surfaces. As
143
demonstrated by (Vetter et al. 2009) water areas typically feature areas void of detected
144
echoes or very sparse returns. It is measured in points per square meter.
145
146
The attributes used for classification are thus: nDSM, Echo Ratio, Sigma0,
Echo Width, and Density.
147
148
3.2.2. Object classification
6
149
First each pixel is classified using the decision tree shown in Figure 2
150
including the threshold values. After the first 2 classes, water and building (candidates) are
151
extracted, mathematical morphology is applied to refine the building results. The pixels not
152
classified are then tested for fulfilling the vegetation criteria. If they are not in vegetation,
153
they are considered to be ground.
154
155
Water is first identified, based on the low point density. As mentioned above,
water has very low backscatter, and often no detected echo.
156
Building objects are distinguished from other objects by height (above 3m)
157
and surface roughness. ER is used to distinguish buildings from tree objects. However, with
158
various shapes of building roof and some buildings being covered by high trees, only ER is
159
not sufficient and would include vegetation in the building class. Thus, EW is used to detect
160
only hard surfaces. Buildings are contiguous objects and have typically a minimum size.
161
This is considered by analyzing all the pixels classified as buildings so far with
162
mathematical morphology. A closing operation is applied first to fill up all small holes
163
inside the buildings, and then opening is performed to remove few pixel detections (“noise”)
164
from the building set. This also makes the outlines of building smoother.
165
ER, Sigma0 and EW are then used to classify trees. The building (and water)
166
mask is applied to classify only pixels not classified before. Finally, all pixels not classified
167
so far are considered ground.
168
169
4. Results & Discussion
170
171
The thresholds for the classification were set manually, based on exploratory
172
analysis of the data sets and on expectation of the objects. The main properties of ER, EW,
173
Sigma0, nDSM, and Density values for Eisenstadt are summed up in Table 1.
174
The results were evaluated quantitatively and qualitatively. Based on the
175
point density characteristic of water region it produces a good result. All the water bodies in
176
the interested area are classified. However, some small parts of the study area where the
177
laser signal could not reach the ground because of occlusion by high buildings, are
178
misclassified. This could possibly be improved with the overlap of another strip.
7
179
While buildings in general can be classified well, very complex roof shapes
180
and walls cause difficulties. It was observed that selecting threshold conservatively the
181
shape of the building is maintained, while its size is reduced slightly.
182
The final classification results were then assessed based on Correctness and
183
Completeness (Heipke et al. 1997). Some buildings, trees and water bodies are digitized
184
manually as reference data. Comparing the results of the automated extraction to reference
185
data, an entity classified as an object that also corresponds to an object in the reference is
186
classified as a True Positive (TP). A False Negative (FN) is an entity corresponding to an
187
object in the reference that is classified as background, and a False Positive (FP) is an entity
188
classified as an object that does not correspond to an object in the reference. A True
189
Negative (TN) is an entity belonging to the background both in the classification and in the
190
reference data.
191
The Completeness and Correction for building, tree, and water class are given
192
in Table 2. It is also illustrated for one building in Figure 4. The two main classes of
193
building and tree feature values above 93%.
194
195
5. Conclusion
196
197
This study used full-waveform LiDAR data to classify urban area, i.e.
198
Eisenstadt city. Four classes are classified: Water bodies, Buildings, Trees and Ground. The
199
computations were executed in OPALS, ArcGIS and Fugro Viewer. Overall, a high
200
accuracy (>93%) could be achieved.
201
Full-waveform LiDAR with its additional attributes is an advanced data to
202
classify urban area. The echo width proofed valuable in classifying vegetation reliably. The
203
other attributes used were Echo Ratio, Sigma0, nDSM, and Density.
204
With the high quality result, LiDAR full-waveform data is suggested to be a
205
data which used for urban classification, urban change detection in the other urban areas in
206
Vietnam in the near future.
207
208
Acknowledgements
8
209
210
The authors would like to thank Vienna University of Technology for
support us the data and software to do this research.
211
212
Reference
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
Alexander, C., Tansey, K., Kaduk, J., Holland, D., Tate, N.J., 2010. Backscatter coefficient
as an attribute for the classification of full-waveform airborne laser scanning data in
urban areas. ISPRS Journal of Photogrammetry and Remote Sensing 65 (5), 423432.
Ducic, V., Hollaus, M., Ullrich, A., Wagner, W., Melzer, T., Year. 3d vegetation mapping
and classification using full-waveform laser scanning. In: Proceedings of the
EARSeL/ISPRS, Vienna, Austria, pp. 211–217.
Heipke, C., Mayer, H., Wiedemann, C., Year. Evaluation of automatic road extraction. In:
Proceedings of the IAPRS "3D Recontruction and Modeling of Topographic
Objects", Stuttgart, pp. 151-160.
Höfle, B., Mücke, W., Dutter, M., Rutzinger, M., Year. Detection of building regions using
airborne lidar - a new combination of raster and point cloud based gis methods
study area and datasets. In: Proceedings of the International Conference on Applied
395 Geoinformatics, Salzburg, Austria, pp. 66-75.
Lemmens, M., 2009. Airborne lidar sensors. GIM International 2 (23), 16-19.
Mallet, C., Bretar, F., 2009. Full-waveform topographic lidar: State-of-the-art. ISPRS
Journal of Photogrammetry and Remote Sensing 64 (1), 1-16.
Mallet, C., Soergel, U., Bretar, F., Year. Analysis of full-waveform lidar data for
classification of urban areas. In: Proceedings of the The International Archives of
the Photogrammetry, Remote Sensing and Spatial Information Sciences, Beijing,
pp. 85-92.
Matikainen, L., Hyyppä, J., Ahokas, E., Markelin, L., Kaartinen, H., 2010. Automatic
detection of buildings and changes in buildings for updating of maps. Remote
Sensing 2 (5), 1217-1248.
Otepka, J., Ghuffar, S., Waldhauser, C., Hochreiter, R., Pfeifer, N., 2013. Georeferenced
point clouds: A survey of features and point cloud management. ISPRS
International Journal of Geo-Information 2 (4), 1038--1065.
Pfeifer, N., Mandlburger, G., Otepka, J., Karel, W., 2014. Opals – a framework for airborne
laser scanning data analysis. Computers, Environment and Urban Systems 45, 125136.
Rutzinger, M., Höfle, B., Hollaus, M., Pfeifer, N., 2008. Object-based point cloud analysis
of full-waveform airborne laser scanning data for urban vegetation classification.
Sensors 8 (8), 4505-4528.
Vetter, M., Hö Fle, B., Rutzinger, M., 2009. Water classification using 3d airborne laser
scanning point cloud. Österreichische Zeischrift für Vermessung & Geoinformation
97(2), 227-238.
Vosselman, G., Maas, H.-G., 2010. Airborne and terrestrial laser scanning Whittles
Publishing, Dunbeath.
Wagner, W., Ullrich, A., Ducic, V., Melzer, T., Studnicka, N., 2006. Gaussian
decomposition and calibration of a novel small-footprint full-waveform digitising
airborne laser scanner. ISPRS Journal of Photogrammetry and Remote Sensing 60
(2), 100-112.
256
9
257
TABLE AND FIGURE CAPTIONS
258
Table captions
259
Table 1. The range of ER, EW, Sigma0, nDSM, Density for Eisenstadt
260
261
Table 2. Accuracy assessments of Building, Tree and Water classes in Eisenstadt region.
262
263
Figure captions
264
Figure 1. Study area
265
Figure 2. Decision tree for the classification. The morphological operations are applied to
266
the binary image of building candidate pixels. Then the second decision tree is used for the
267
classification, focussing only on tree and ground class. The threshold values apply to the
268
Eisenstadt area.
269
Figure 3. Urban full-waveform classification in Eisenstadt.
270
Figure 4. (a) Ground truth data; (b) classified result; (c) accuracy assessment
271
10
272
TABLES AND FIGURES
273
Table 1.
ER [%]
Eisenstadt 4,2-100
EW [ns]
0-29
Sigma0 [m]
0-19
274
275
11
nDSM [m] Density [pt/m2 ]
-1.52 – 39 0-72,4
276
Table 2.
Building
Tree
Water
Comp
97.3%
97.8%
89.0%
Corr
96.0%
93.9%
90.7%
Quality
93.7%
92.0%
81.6%
277
278
12
279
280
Figure 1.
281
282
13
283
284
Figure 2.
285
286
287
288
14
289
290
291
Figure 3.
292
293
15
294
Figure 4.
295
16
Download
Related flashcards
Create Flashcards