Transcript Document

Shibasaki-Lab, the University of Tokyo.
Extraction of Urban Area
Using ASTER Imagery and the Existing Land Cover Data
Ayako TANAKA*, Koki IWAO**, Ryosuke SHIBASAKI*
Center for Spatial Information Science, The University of Tokyo*,
Grid Technology Research Center, National Institute of Advanced Industrial Science and Technologiy**
Objective
Backgrounds
Social background !!
Problems !!
Urban Area Mapping using ASTER images
and the Existing Land Cover Data
Riyadh, Saudi Arabia
Urban population growth
MOD12
GRUMP
ASTER image
Urban extents Grid *1
Land cover *2
Important points in developing urban mapping method:
Urban expansion
1, Global !!
: urban area
GRUMP/ urban extent
http://bet.sedac.ciesin.columbia.edu/gpw/
2, Automatic and Simple
3, Better Resolution and Accuracy
: urban area
: non-urban area
: vegetation
Overestimate!!
Climate change
Coarse!!
To analyze and evaluate effects for human activities
More
accurate!!
We need New
ASTER
+
Global Urban Area Map !!
: urban area
MOD12
http://www-modis.bu.edu/landcover/
GRUMP urban extents Grid
(city lights)
*1 GRUMP urban extents grid: Urban extents grid is a data set resulting from GRUMP (Global Rural-Urban Mapping Project) include a 30 arc-second land area grid showing urban areal extents worldwide
produced by the Center for International Earth Science Information Network (CIESIN) of the Earth Institute at Columbia University.
+
*2 MOD12: This is one of the Land cover products generated by BOSTON University using MODIS Data include a 30 arc-second land area grid. Urban areas are one of the land cover category.
MOD12 (seasonal change)
Methodology
Images in each Process ：Case study of Asahikawa
Urban area
(detected by
visual
interpretation
)
GRUMP/urban
extent thin2
thin1
urban
urban
①ASTER image
urban
ASTER classification
⑦MOD12/ urban
④GRUMP/ urban extent
MOD12
buffer
urban
Overlay・Region Segmentation
②ASTER classification
Classify the urban area
using unsupervised
classification method
for the automatic data
processing
⑤GRUMP/ thin
Thinning is a
morphological operation
that is used to remove
selected foreground
pixels from binary
images (3km, 5km)
“Probability of each segment as urban” =
⑧MOD12/ buffer
Buffering is a
morphological operation
that is used to remove
selected background
pixels from binary
images (2km)
Urban area detected by visual interpretation (pink)
Area of each segmentation (gray)
Calculate “probabilities of
each segment as urban” in
6cities (Asahikawa, Oita,
Manaus, Hailar, Dubai).
・Average ≧ 80% ：
classified as urban
③ Results of visual
interpretation
⑥Overlay/
GRUMP&Result of
visual interpretation
⑨Overlay/
MOD12&Result of
visual interpretation
・ Average ＜ 80% ：
classified as others
Probability of each segment as urban
←⑩ Urban Map
（Detected by this method）
Result
Accuracy
分類
100
※Apply this method to 4 cities in different climatic zones
① Temperate Zone：Asahikawa
② Desert Area：Dubai, UAE
③ Steppe Area：Hailar，China
④ Tropical Zone：Manaus, Brazil
80
MOD12 /urban
Rate (%)
Hit的中率(%)
60
Results of urban
classification
asahikawa
oita
manaus
ulan
hailar
dubai
分類結果
Result of
using this method
40
GRUMP/
urban extent
20
0
0
40
60
80
100
Detection
Rate(%)
検出率(%)
Results of visual
interpretation
※test area : ①Asahikawa (Japan), ②Dubai(UAE),
③Hailar(China)， ④Manaus(Brazil), （⑤Oita(Japan)，⑥Ulaanbaatar(Mongolia)）
Conclusion
Study Area
⑥ ③
②
④
20
①
⑤
Urban areas are tend to be overestimated in the results.
Detection Rates of the results : 60 – 95 % (Higher than that of MOD12)
Hit Rates of the results : 50 – 70 % (Higher than that of GRUMP/ urban extent)
This method improves
•detection rates compared with MOD12
•hit rates compared with GRUMP/ urban extent
・ Detection Rate(%) = Urban area (overlap) / Urban area (detected by visual interpretation) ×100
・ Hit Rate(%) = Urban area (overlap) / Urban area (detected by this method) ×100
Detection Rate（検出率）：
The rate of urban area detected by visual interpretation in the urban area in the results
Hit Rate（的中率） ：
The rate of urban area detected by visual interpretation in the urban area in the results
Future works
・Apply this method to various cities and assessing the accuracy
・Generate a global urban map using this method
・Classify small villages accurately using
–Gazetteer :a dictionary of place-names *
–slope data
* What is a Gazetteer?
A Gazetteer is a geographical dictionary. Instead of including lists of words, it includes lists of places. These places can be settlements
(cities, towns, villages) or geographical features (hills, rivers, regions, parks, tourist attractions etc.). We have extended the concept of a
Gazetteer to include famous people and family names.