Dowman, Ian 
  
3.1 Image capture 
The imaging unit (Figure 2) was developed as part of a project called HAZMAP, which was carried out at the Department 
of Geomatic Engineering of University College London. A main goal of HAZMAP was to minimise human intervention 
during the image acquisition in hazardous environments (Chapman ef al., 1994). The images are captured in digital form 
and are displayed online on the screen of a computer. The image acquisition is performed using an off-the-shelf Pulnix 
TM526 CCD camera which is mounted on the motorised Geotronics 510 SN servo-driven Geodimeter. 
  
Fig. 2: Image acquisition unit. 
The concept of camera-on-theodolite was originally introduced by Huang and Harley (1990) but is used here for the 
capture of terrestrial images from building facades in order to speed up the process of texture creation. The system is 
quite simple and requires no extensive training for its setup and the corresponding operations. At each imaging location 
(station) a set of image tiles that cover a field of view of 360? (horizontally) by 90? (vertically) is taken. Once captured, 
the images along with the theodolite angular readings are stored to the hard disc of a PC computer. The process of 
recording theodolite angles for each image is called indexing. Indexing the images considerably simplifies the task of 
image registration which is required for the processing of textures. This is because, at any station the angular location of 
images with respect to each other and the theodolite centre is known. Therefore, by knowing only the exterior orientation 
of stations, the relative orientation of all images in the database with respect to each other is determined. Using the index 
data the reference file is created which mainly defines the theodolite angles, dimensions, and the location of each image 
on the hard disk. 
3.2 Registration of CCD images 
The registration of images includes the calibration of the data capture unit and the localisation of the imaging stations. 
Within the camera calibration stage, not only the characteristics of the CCD camera lens but also the parameters relating 
the geometry of the CCD chip to the rotation centre of the Geodimeter are determined. The images used to calibrate the 
instrument are usually taken before the image acquisition is started. Having done the calibration, the images are acquired 
and the stations are localised. Localisation refers to the process in which the exterior orientation elements of the theodolite 
centre at each imaging station are identified. Central to the calibration and localisation is a mathematical model relating 
the CCD and object coordinates of a point using an extended form of the Collinearity Equations. Details of this model 
can be found in Varshosaz (1999). The outcome from the calibration and the localisation stages are the camera and the 
station files respectively. 
Once the images are captured and the reference, camera, and station files are created, the TID is formed and can be used 
to process the texture of a given face using the ATPT. 
4 THE AUTOMATIC TEXTURE PROCESSING TOOL 
Given the coordinates of the corners of a building face, the ATPT searches the reference file and automatically extracts, 
rectifies, and mosaics the image portions covering the face. The output is the texture of the given face. Within the 
current implementation of the ATPT the faces are assumed to have a rectangular shape. This assumption simplifies the 
development of the tool and is acceptable for most building facades in urban areas. In addition, the creation of textures for 
building parts having a different shape follows mainly the same principles as the rectangular faces. The texture processing 
is carried out for all building faces individually, for each of which a texture is created. 
Figure 3 shows how the texture processing is carried out. To from the texture of a face, the 3D coordinates of its corners, 
extracted from the NFR, are passed to the ATPT that uses the camera calibration, localisation, and reference data to find 
the images which cover the face of interest. Based on a given spacing interval a 3D grid, the points of which define the 
elements of the texture, is fitted to the face. 
  
182 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B5. Amsterdam 2000. 
  
  
The 
For 
COO 
finz 
fact 
4.1 
For 
is a 
Th 
par 
dis 
of 
the 
COI 
the 
an 
sta