quence of 
id. ordered 
'he field of 
ven image 
d camera. 
xment from 
adapter. For 
a randomly 
that is based 
pping zone. 
le geometric 
The use of this types of panoramas for exact geometric 
reconstruction is limited. Nevertheless they provide efficient 
overview capabilities for the content, the current use or just 
the local properties of a room or an environment. Due to low- 
cost investments for such programs (less than 500€), and the 
provision of easy-to-use software plug-ins, they are widely 
distributed. As an example, in the field of facility 
management digitised 2-D building plans are used to register 
the position of panorama stations. By simple navigation and 
clicking the local environment can be observed using the 
corresponding panorama views. Additional information on 
panorama images, virtual reality and presentation of 3-D 
models can be extracted from Hóhle (1998), Düppe (1998), 
Pomaska (1998) and Hóhle, J. & Pomaska, G. (1999). 
3. GEOMETRICALLY EXACT PANORAMAS FOR 
3-D DATA ACQUISITION 
In the following paragraph the calculation of 
photogrammetric panoramas is discussed. For our 
experiments a calibrated digital camera Fuji FinePix SIPro 
with 15mm fish-eye lens has been used. With this 
camera/lens combination it is possible to acquire a panorama 
by only 6 individual images with appropriate overlaps. 
3.1 Calculation of distortion-free images 
Given the parameters of interior orientation of the camera 
(calibration data) the acquired images can be rectified to 
distortion-free images, i.e. providing images of an "ideal" 
camera that is free of distortion and image deformations. 
Hence, the transformation of object information into image 
space is strictly performed under the laws of central 
projection. 
Fig. 6 shows clearly the effect of lens distortion of the 15mm 
fish-eye while, in the distortion-free image, linear object 
structures are also displayed as linear features. The metric 
size of the corrected image is not equal to the original one, 
i.e. the black areas correspond to image positions outside the 
actual format of the original image. 
  
Th 
i 
HE 
  
  
  
  
  
  
  
  
  
     
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Fig. 6: left: original image; right: calculated distortion-free image 
Bild(1) Punkt(1 
  
) 
Fig. 7: Measurement of identical points in the overlapping zone 
The camera is mounted on a tripod. In order to determine the 
imaging directions of each picture, tie points are measured 
within the overlapping zone of adjacent images. Since at least 
small tilt and roll angles are existing that have to be calculated 
as part of the orientation parameters of the panorama, a 
minimum of two tie points is necessary for each overlapping 
zone (Fig. 7). 
3.20 Mathematical basics 
Prior to the actual orientation process the excentricity of the 
objektseitigem perspective centre of the camera with respect to 
the (vertical) rotation axis of the mounting device has to be 
determined. It is assumed that the excentricity remains constant 
of the acquisition period of 6 images. Fig. 8 shows the 
geometry of the excentricity vector. 
The correction values dx'(ey) and dx'(ex) are applied to the 
image observations. In this way "ficticious"image observation 
are generated whereby the imaging geometry is related to the 
rotation axis of the tripod. This leads to the following 
equations: 
Correction of measured image coordinates x ' 
x' 2 x'+dx'(ey) + dx' (ex) (1) 
' 
where dx'(ey) = LOCO. and dx'(ex) = £k . ox (2) 
Y+ey y 
The equations show the vertical distance Y from image to object 
point which is not available due to missing knowledge of the 
distances of all object points to the camera station. However, 
since the diameter D of pasted target points is known, the 
measured image diameter d lead to a local image scale in each 
point. Hence, Y can be written as 
—183—