:hers 
oach 
city 
good 
nage 
t the 
‘hich 
irdly 
ality 
2 can 
) test 
using 
dings. 
. The 
ation 
large 
ning 
nsing 
logic 
| it is 
odels 
ethod 
et the 
: and 
ligital 
v 3D 
y the 
| has 
prises 
L RS 
1mera 
is the 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B-YF. Istanbul 2004 
  
carrier of RS device which is made mainly of glass fibre 
reinforced plastic, and it has a lighter weight but a higher 
intension. The tele-controlling system is used to control the 
flying state and all devices working normally. Fig.1 shows the 
outline of UAVRS-II. 
    
Fig.] UAVRS-II RS System 
In traditional aerial photograph, because the photo is taken 
nearly vertically to the earth surface, we can only get the height 
and top texture information of the main part of building, and 
lose plenty side face information. That can't meet the demand 
of acquiring subtle city information. Unlike the traditional 
aerial photograph, we should photograph the object building 
obliquely (Fig.2) when we want to get the information as much 
as possible. From the oblique photo, we can derive various 
types of information including the height of every main part and 
small part, the top texture and the side texture. 
  
  
  
  
  
  
Fig.2 Oblique Photograph 
UAVRS-II can do remote sensing in the height range from 50 to 
4,000 meters. Although we can acquire the high resolution 
image with the vehicle flying in a low height, there may be a 
danger of bumping tall buildings and the photo could only 
cover a small region. And if the vehicle is flying very high, the 
resolution of photo will be low grade. From lots of tests, we get 
the UAV images on various shooting conditions, and at last we 
draw the conclusion that 45? and 300 meters are the ideal 
shooting angle and height (Fig.3). 
Although UAVRS-II has many good characteristics such as low 
cost, high resolution, flexibility and etc, it also has some 
disadvantages. Because it has a very light weight, UAVRS-II is 
much more influenced by outside environment such as the wind 
and it is difficult to control the vehicle photographing on a same 
pose. Therefore we have relatively much trouble in putting the 
images together. In the paper, we put forward a new method to 
do surveying on a single image, so we successfully avoid the 
limitation of UAVRS-II. 
101 
"Oe 
  
00€ 
  
Ground Building 
  
  
  
  
Fig.3 The Ideal Photographing Pose 
3. SURVEYING ON A SINGLE IMAGE 
In traditional photogrammetry, we usually use an image pair to 
reconstruct the 3D model of spatial object. But to some objects 
which have regular shapes such as building, because there are 
certain geometric relations among the out lines, for example as 
showed in Fig.4, Top line is parallel to Ground line and vertical 
to Side line, we can reconstruct the 3D model of building using 
these geometric attributes. Many scholars distributed all over 
the world have done researches on such subject (T.E. Chen, 
Chen Zen, Frank A.van den Heuvel, Zuxun Zhang, et al). 
In the paper, we design and implement an single image 
surveying method, by which we can get both the height and 
texture of buildings. The whole process could be divided into 
four steps. In a single UAV image, at first we get the inner 
elements (x; yy) and f by checking up the camera, secondly we 
divide the parallel lines in buildings into three groups which 
parallel X, Y and Z coordinate axis respectively and compute 
the three joint points, thirdly calculate the three angle elements 
A,aandkto implement the image relative orientation, fourthly 
compute the scale between reality and image using the real 
coordinate information abstracted from 2D GIS databases, then 
on the basis of all the above parameters we can get the real 
height and the ortho-texture of buildings. 
a. Checking the camera 
The purpose of camera checking is to get the inner elements 
including (x; yo) and / and the lens aberration parameter K,. In 
the tests we used Nikon D100 digital camera, and the checking 
results is showed in the following formula. 
#, 1209 
T «3261233 
K, 2 -129e - 008 
The result unit is pixel, and by multiplying 7.8um which is the 
width of a pixel we can transform it to millimeter. 
The following is the model to rectify lens aberration.