531
DEVELOPMENT OF THE REVISING METHOD FOR NATIONAL DIGITAL LAND-USE INFORMATION
USING REMOTELY SENSED DATA
Yasushi Shimoyama, Minoru Akiyama, Tokio Mizuno, Takashi Hayashi
Geographical Survey Institute (GSI) of Japan
ABSTRACT
National Land Agency (NLA) and GSI have been prepared Japanese national land information in digital
format since 1974. As far as land-use information is concerned raster image data of 15 categories with pixel
size of approximately 10m by 10m were collected from 1:25,000 scale topographic maps. Those data are
expected to be revised periodically. However, besides budgetary reasons, we cannot fix the revision date by
this method because original topographic maps are revised in sheet by sheet and not simultaneously.
Therefore, NLA is planning to revise the data using satellite image data, and GSI has been studying on
developing the revising method. The method is not fully automatic image analysis but quite practical one which
integrating visual image interpretation techniques into digital analysis. The outline of the method is as follows.
For registration of satellite images onto maps, visual matching between rasterized map and satellite image on
color display is adopted which is quite simple and easier than GCP collection. Classification is mainly done
by manually determined decision boundaries on two dimensional feature plane. In this case, decision boundary
is drawn by hand with the help of information for cluster distribution and training data distribution on the same
two dimensional feature plane. Moreover, decision boundary can be edited by hand with referring to the
classified result. Classified result is compared with the old land-use data so that we can check if the direction
of landuse change is reasonable or not. Final result is arranged after checking and correcting the classified
result by image editing basis with the help of photo interpretation technique.
In the paper the concepts, methodology and a case study of this method will be presented.