COMPUTER ASSISTED INFORMATION EXTRACTION FROM SATELLITE IMAGES
FOR UPDATING NATIONAL LAND USE INFORMATION DATA BASE IN JAPAN
K. Cho, M. Yoshimura & S. Takeuchi
Remote Sensing Technology Center of Japan
S. Murai
University of Tokyo, Japan
C. Otsuka
Geographical Survey Institute of Japan
K. KamadaX
National Land Agency of Japan
Commission IV
ABSTRACT:
A personal computer based image interpretation system called CASYII(Computer Assisted
System for Image Interpretation) has been developed. In this system, human image inter-
pretation technique is highly supported by it's various graphic functions, and allows
users to update land use information effectively by comparing old land use data and
latest satellite images on the display. Since FY 1991, the National Land Agency of Japan
has been using this system for updating the national land use information data base from
latest satellite images. This paper describes the outline of this project, and explains
how the human image interpretation technique is combined with digital image processing.
KEY WORDS: SPOT, HSI, Image Interpretation, Land Use Data, CASYII
1. INTRODUCTION 1.2 Limitation of Point-wise
1.1 Digital National Land Information clossification
Landsat TM and SPOT data suggested us
possibilities of extracting detailed land
eover information from satellite data.
However, at the same time, limitation of
applying traditional point-wise classifi-
cation methods, such as the maximum like-
lihood classification, to the high spatial
resolution data have become clear. One of
the main reason of this is that the con-
ventional land use items can not always be
represented by particular spectral char-
acteristics of satellite data. Fig. 1
shows the spectral characteristics of the
each land use item area which were updated
by image interpretation of a SPOT/TM
composite image.
Since 1974, under the cooperation with
Geographical Survey Institute and other
agencies, the National Land Agency of
Japan has been collecting and updating
various land information in digital form
called the Digital National Land Informa-
tion. This information include but not
limited to topographic data, geological
data, climate data, land use data, admin-
istrative division data. Most of the data
are produced and updated by in situ inves-
tigations, air photo interpretation,
topographical map interpretation etc. As
the conventional survey method takes lots
of time and cost, the National Land Agency
had been investigating the possibility of
using satellite remote sensing technology RA
for updating the data.
Among the Digital National Land Informa- ;
tion, the land use data was expected to be
one of tbe most suitable data to be updat-
ed from satellite images. The mesh size
of the original land use data was 10m, and
100m mesh data for public use were creat-
ed from the 10m data with majority vote
method. The original land use data were
consisted of 15 items, which are shown on
Table 1 as "old land use items". In order
to reduce the time and cost for updating
the land use data, under the contract of :
National Land Agency and Geographical T (see Table 1 for item number )
Survey Institute, Remote Sensing Technolo-
gy Center of Japan(RESTEC) has been in- 0 + + i I + t > G
volved in studying the possibility of
using satellite data for it. Fig. 1 Spectral characteristics of the
each land use item area which
were updated by image interpreta-
* Moved to the Geographical Survey Insti- tion of a SPOT/TM composite image.
tute of Japan in April, 1992. (see Fig.3)
so +
122
This r
ing po
data
Textur
the so
These
of ini
approa
renewa
2.1 De
In or
venti
author
updati
tion :
tal iT
conce
summal
(1)Ut:
te:
(2)Ut.
(3)Di'
* P
