THE INTEGRATION OF REMOTE SENSED IMAGES AND DTM DATA 
Weihsin Ho 
Department of civil engineering 
National Chiao Tung University 
1001 Ta-shuei Rd. Hsinchu Taiwan, R.O.C. 
Liang Huei Lee 
Department of surveying and mapping 
Chung Cheng Institute of Technology 
Ta-hsi Toayuang Taiwan, R.O.C. 
ABSTRACT 
Images obtained from remote sensors and information of DTM are the 
Especially, three dimensional image-based display has better 
vector-based one. The integrated usage of these two kinds o 
development in GIS. In this paper, we will discuss the 
images and DTM data. The topics include (a) to capture image-based terrain features which can be 
displayed alone or with remote sensed images togather, (b) to do 3D projections of remote sensed images 
for landscape visualization, and (c) to simulate a stereopair from a single image. The principle of 
processing is to satisfy the condition of that each pixal on images should match with a certain height 
within DTM. The resolution of these two kinds of data is different, therefore in general, the DTM data 
should be processed by interpolation to subject to those of images. In the paper, we also propose the 
algorithms for different situations such as terrain slope, aspect, ridge lines, valleys, streams and 
shadows etc. 
important data sources for GIS. 
visualization effect than three dimensional 
f information is the important direction of 
processing of integration of remote sensed 
Keyword: GIS, DTM, landscape visualization 
I. INTRODUCTION II. DTM DATA PROCESSING 
GIS is a information system of spatial data. It 2-1 FILTERING 
is widely applied to the related realms of 
resource management and planning [15]. In GIS, DTM data obtained from profile scanning or image 
a lot of land surface information is obtained correlation method contains dynamic scanning 
from remote sensed images and DTM data. noise and measurement errors. This process is a 
Furthermore, images is also the important data smoothing procedure to reduce noise and error. 
source when spatial data is updated. At the We can delineate the dynamic scanning using the 
present time, many GIS are vector-based following differential equations [1]: 
systems. Hence, the display is vector linework 
in such systems. However, linework display has Z -Z*r1Z«r2 Z (1) 
worse visualization effect than raster-based Z «dz . Zi.1-7i.1 (2) 
image display. And there is a lot of dy 2Ay 
information of the terrain features that can be = 2 
deri ; 7 -02z.27:1:240:*2,., (3) 
erived from raster images. Therefore, how to S EIU T 
combine these two kinds of information is one dy Ay 
of the important trends in GIS community. where A Yis DTM interval which can be normalized 
There are many applications which need the to 1, 
integrated information of remote sensed images ri,rpare filtering parameters, and 
and DTM data. Basically, remote sensed images Zi are real height and measured height 
can be processed with DTM data in the * respectively. 
followings : (1) ortho-rectification for the 
registration with map coordinates, (2) raster- 
based extraction of terrain features, (3) Because real height, Z7 is difficult to obtain, 
the 3D projective transformation, and (4) it can be replaced by the height from the 
simulated stereopair from a single frame of neighboring profiles with self-calibration 
images. The principle of processing is to filtering, i.e. 
content that each pixel on images should match 
with a certain height within DTM. The 7-7 - 2Z,i- Zi: Z1 (4) 
resolution of these two kinds of data is A MU P 
different, so in general, the DTM data should 
be processed by interpolation to coincide with 
those of images. In this paper, we introduce : 
where j is a number of profiles (j=1,2,...,m), 
and i is a resampling point in the jth 
profile. Combining eg.(1) to eq.(4), we can 
such as the extraction of terrain slope, solve the filtering parameters using the least 
shadows etc. Finally, 3D transformation models Squares method based on 3 to 5 profiles. Then 
and stereopair simulation have processed by the height after filtering can be calculated 
means of methods of computational vision. with respect to the parameters. 
some preprocessing of DTM data first. We also 
propose the algorithms for different situations 
902 
as 
ads" Aha Baud? Bath 
be £N ds B 
t m o. €