LANDSLIDE FEATURES INTERPRETED BY NEURAL NETWORK METHOD USING A
HIGH-RESOLUTION SATELLITE IMAGE AND DIGITAL TOPOGRAPHIC DATA
K. T. Chang* * and J. K. Liu”
* Dept. of Civil Eng., MUST, No. 1, Hsin-Hsing Rd., Hsin-Fong, Hsin-Chu 304, TAIWAN. ktchang@must.edu.tw
^ Dept. of Civil Eng., NCTU & ERL, ITRI, Hsin-Chu 311, TAIWAN. JKLiu@itri.org.tw
Commission VII TS WG VII/S
KEY WORDS: Landslides, Earthquakes, Hazards, Quickbird, LIDAR, Neural
ABSTRACT:
Landslides are natural phenomena for the dynamic balance of earth surface. Due to the frequent occurrences of Typhoons and
earthquake activities in Taiwan, mass movements are common threatens to our lives. Moreover, it is a common practice for the
agencies of water reservoirs in Taiwan to make a reconnaissance of the landslides of the watershed every 5 to 10 years for the
purpose of conservation. It is found that the application of aerial photo-interpretation technique for this purpose has been recognized
as an effective approach since 1970s. However, an efficient and automatic interpretation scheme has never been established.
Therefore, two issues are to be resolved for creating a useful and timely landslide database, i.e. the consistency of the sub-datasets
and the completeness of the coverage. As the manual interpretation and automatic recognition are compared, the former is a
practical and operational method, but the result it derived is largely dependent on the professional background of interpretation
operator.
[n this paper, the interpretation knowledge is quantified into recognition criteria. Multi-source data, e.g. a Quickbird satellite image,
DTM reduced from a LIDAR data, road and river vector data, are fused to construct the feature space for landslides analysis. Then,
those features are used to recognize landslides by a multilayer perceptron (MLP) Neural Network Method. The extraction result is
evaluated in comparison with the manual-interpretation result. The experiments indicate that the conducted method can assist
landslide investigation efficiently and automatically. Moreover, th
e ANN method is better than some statistic classification methods,
e.g. Maximum Likelihood method, due to its adaptability for multi-resource data and no predefined assumption.
1. INTRODUCTION
Landslides are natural phenomena for the dynamic balance of
earth surface. The potential or intrinsic factors of landslide
include geological and morphological factors and the external
or triggering factors include earthquake, climate, hydrology,
and human activities. When the geology is highly fractured and
landforms are in high relief. In addition, the frequent
earthquakes and heavy rainfalls are together imposing further
stress to the earth to break the balance of the nature. And, thus,
mass movements such as landslides, slumping, and mudflows
take places.
1.4 Motivations
Moreover, it is a common practice for the agencies of water
reservoirs in Taiwan to make a reconnaissance of the landslides
of the watershed every 5 to 10 years for the purpose of
conservation. It is found that the application of aerial photo-
interpretation technique for this purpose has been recognized as
an effective approach since 1970s. However, an efficient and
automatic interpretation scheme has never been established.
Therefore, two issues are to be resolved for creating a useful
and timely landslide database, i.e. the consistency of the
datasets and the completeness of the coverage. As the manual
interpretation and automatic recognition are compared, the
former is a practical and operational method, but the result it
derived is largely dependent on the professional background of
interpretation operator.
It is usually taking a long time to make a large-scale and real-
time mapping of landslides after a torrential rainfall. The first
general mapping of landslides in Taiwan was conducted by Soil
and Water Conservation Bureau in 1982-1989 and a landslide
* Corresponding author. Assistant prof., Dept. of Civil Eng., MUST, N
ktchang(@must.edu.tw, +886-3-5593142 x 3297
map of Taiwan in a scale of 1/50,000~1/100,000 (COA, 1991).
In 8 years of survey, there were more than 10 times of torrential
rainfalls and 100 times of earthquakes and new balance of the
nature took time and time again. In reality, to map all the
landslides in one time is not feasible. And, it is understandable
that the difficulties of obtaining a survey with completeness of a
whole Taiwan coverage. It has been a common practice to
interpret aerial photographs by visual inspection of an expert
geologist. It is a time consuming task. Therefore, the purpose
of this study is to implement the human rules and quantifies the
criteria to install an automatic system by a back-propagation
Neural Network Method.
1.2 Overview and References to related works
Landslides cause approximately 1000 deaths a year worldwide
with a property damage of about US$4 billion, and pose serious
threats to settlements and structures that support transportation,
natural resource management and tourism. In many cases, over-
expanded development and activities, such as slope cutting and
deforestation, can sometimes increase the incidence of landslide
disasters. Recent development in large metropolitan areas
intrudes upon unstable terrain. This has thrown many urban
communities into disarray, providing grim examples of the
extreme disruption caused by ground failures (Singhroy &
Mattar, 2000).
Aerial photography has been used extensively to characterize
landslides and to produce landslide inventory maps, particularly
because of their stereo viewing capability and high spatial
resolution (Liu, 1985, Liu, 1987). However, the conventional
photo-interpretation is a time-consuming and costly approach
(Liu et al., 2001).
o. 1, Hsin-Hsing Rd., Hsin-Fong, Hsin-Chu 304, TAIWAN.
574
Intern
Satell
param
geom
Multi:
veget:
morpl
et al.,
For 1
featur
sensir
for ex
image
would
contra
veget:
the ele
and ty
and st
a gen
mostl:
ct al,
Resol
used :
using
expert
impro
(Kojir
Artifi
in m:
appro;
autom
been 1
orgag
prove
too pc
more |
24.1
Indivi
locati
depen
fall, s
materi
mech:
metho
survey
Groun
take p
make
after a
Photo;
and
autom
geolog
envirc
landsl
judge:
landsl
The a
the ap
and sj
rainfa|
length
