(7)The panchromatic image of 1, 2.5 and eight meters and the
multispectral images of 4, 10 and 16 meters were generated.
Consequently, the Pa/Mu composite images of 1/4, 2.5/10 and
8/16 meters were prepared.
2.3 Interpretation
The geographic information to be interpreted is determined from
the legends of actual thematic maps shown in Table 2, because
most of GIS data are generated from them. In case of a geological
map, rock types were not interpreted because it was reported that
geological classification was very difficult in the area with plants.
Some unmapped items that are not specified in the legends but
recognized clearly in the images like inter-change and shadows
were also interpreted.
The interpretation was carried out using the generated simulated
images and actual satellite images of AVNIR. Both multi spectral
image and Pan-sharpened image of AVNIR were interpreted. The
band 2, 3 and 4 of AVNIR images were used for multi spectral
bands, while the simulated images are in natural color.
The interpretation was carried out under the following conditions.
(1) The interpretability is dependent of the image scale. The inter
pretation was performed on the display at any scale.
(2) An expert with some experiences would interprets.
(3) Absolute attributes(ex. university or college) were not
necessary to interpret.
(4) Only images would be interpreted . Existing maps would not
be used as reference or preliminary knowledge.
(5) The interpretability was ranked into 4 levels; l)derivable auto-
matically(with computer), 2)well interpretable(by human
interpretation), 3)poorly interpretable(to some extent),
4)not interpretable.
Table 2 Used maps
Scale
type of map
1:10,000
administration map
1:25,000
topographic map
1:25,000
land use map
1:50,000
geological map
3. RESULTS
The simulated images are clearer than real AVNIR images on the
same scale. It seems that point spread function of simulated
images is better than that of a real AVNIR image. The AVNIR
image has band-to-band registration error to sonie extent. That is
also the reason for image quality.
Table 3 shows the summarized results of interpretation. Most of
objects were easier to interpret in proportional to the spatial
resolution. Forests and river regions are derivable from ADEOS
images, because near infrared bands were available.
4. CONCLUSION
Inr order to evaluate what kinds of geographic information can be
interpreted from high resolution satellite data, the following tasks
were conducted..
(1) The high resolution satellite data were simulated from color
air photographs. The atmospheric effects and data
compression were applied to the generation of simulated
images.
(2) The geographic information extracted from satellite images
was selected from the legends of existing maps with respect to
map scales.
It was proved through the study that the interpretable information
would increase tremendously after the realization of spatial
resolution of ALOS. The following studies should be carried out
for more accuracy.
(1) The development of algorithm for the generation of more
accurate simulated images.
(2) The check of pranimetric accuracy of the geographic inform
ation derived from satellite data.
(3) The analyses of the interpretation level and the suitable image
enhancement for interpretation.
REFERENCES
[RESTEC, 1995] RESTEC, 1995, ‘Report on the Realization of
ALOS Mission Requirements’
[Kawata, 1997] Y. Kawata, et al., 1997, ‘The estimation of
Atmospheric Point Spread Function and Removal of
Atmospheric Effects in the Frequency Domain’, Current
Problems in the Atmosphere, Deepak Press
PI-4-2
;
H
»
Wi
m
ii
mm
