on of
n to
005;
,
netry
| and
lition
’hree
d In
three
tures
se in
ption
ition
ition
ction
are
"d in
gure
) nm
t has
WIR
other
use
num
and
1g in
this
ntly
oily
| the
tion
erial
gets
ie to
face
and
1 of
the
ions
ased
tion
ood
The
ct to
e of
i of
suit
the
Reflestance[%]
FETESCOITE
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B7, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
Spectral Library Plots
rT TTT
A À
05 67 09 1i
PEUT ETES
Wavelength
13
de FAST
Nen
2000
1500
1000 F-
stl A tt by
0.5 0.7 2.9 1.3 1. avaagth 1.7 1.8 2. 2.3
Figure 8. VIS2 and SWIR ratio ranges.
Asphalt condition spectro
2000 + eC ur
- a
- fu
1500F 4
- / ^ i ct an hm X
L^ E gU
1000 e
^ Lr
7 1.7082 um -1.7323 ym
Figure 9. Hydrocarbon absorption band ranges.
Figure 10. Classification result for the whole study area.
MUnclassified
@Bad—asphalt
termediate-asphalt
@ Good—asphalt
EMasked Pi
Class Image li a
2200} UD NIC ge
2000r
1800
^
1600f /
Reflectance
1400} ;
1200F
1
s se sat 20
Figure 11. Assessment of classification result on SteinbeisstraBe.
Even though a comprehensive evaluation is not yet finished, the
spot checks supported by field visits indicate a high potential of
the approach for identifying roads with good, intermediate and
bad surface condition.
4. CONCLUSION
This study focuses on two main purposes, namely identification of
road surface materials and investigation of different conditions of
asphalt. The classification results show that the SAM
classification based on regions of interest is helpful for
discriminating road surface materials. Regions of interest
represent the mean spectrum for an area of interest and thus take
into account variations in spectra of materials due to age or usage.
Additionally, combining mean and standard deviation spectral
functions is helpful for distinguishing asphalt, concrete and gravel.
This is possible since asphalt has a lower mean value compared to
the other two materials and concrete has a lower standard
deviation than gravel over the wavelength range of 619.9nm-
1323.7nm. In terms of condition determination for asphalt roads, it
is observed that the mean function gives reliable results with good
success in identifying roads with good, intermediate and bad
surface condition. This is because the spectra of different
conditions of asphalt differ significantly in albedo. From the
research, it was observed that hydrocarbon absorption bands are
useful in surface material condition investigation. In particular the
wavelength range 1.7082 pm to 1.7323 um are suitable for
identifying different states of asphalt.
It is observed that the number of unclassified pixels in the results
presented in this paper is generally significant. Therefore, more
research should be done to improve the methods adopted and thus
reduce the number of unclassified pixels. Additionally,
hyperspectral data with better spatial resolution should be used.
