WW -— e 
melt "rt: ot 9 
yx 
  
  
Dowman, fan 
  
  
    
  
Face corners 
  
    
| spacing 
interea 
  
  
  
  
Fig. 3: Flowchart of texture processing 
The basic goal of the ATPT is to define the pixel values associated to the grid points which are processed one at a time. 
For each point the reference file is searched and an image with the most orthogonal view over the point is selected. The 
coordinates of the point on this image are then computed using the Collinearity Equations. An image interpolation is 
finally carried out to estimate the pixel value of the point. The estimated value is then written to a new image which is in 
fact the texture of the face. 
4.1 Finding the best image 
For each 3D point, an image has to be found in order to extract the pixel value of the point. Finding the image (Figure 4) 
is accomplished using the following information: 
e H and V: the horizontal and vertical angles, recorded by the theodolite, for the centre of the images; 
e Camera calibration parameters; 
e Station locations and orientations. 
  
  
  
3p Station fi | Camera der à : 
| Coorüinates | | Shon foe | | calibration file | | Inder fir | 
mo T 
Compute H and Kead H and V 
V of pant P of an image 
  
     
   
  
  
  
  
  
  
  
Calculate Caiculate 
coordinates of p coordinates of i 
  
  
  
Fig. 4: Find image process Fig. 5: Distance between the image and point vectors. 
This data is used to relate the image and object points via the collinearity equations. The process relies on the computation of a 
parameter which is used to rank the images; the image with the most appropriate parameter is then selected. This parameter is the 
distance between two vectors with a size equal to 1. Both vectors start at the CCD camera centre, but one of them points to the centre 
of the image while the other refers to the point. As shown in Figure 5, the distance ip is used to rank the images. Points p and i are on 
the vectors pointing to the 3D grid point and the image centre respectively. Once a 3D point is passed to ATPT, the coordinates of p are 
computed. Then, the angles of the images at the station, determined in the previous stage, are read one set at a time. For each image, 
the coordinates of i, and consequently the distance ip is calculated. Every time a new image is introduced to ATPT, its ip is computed 
and compared against the one obtained for the previous image. If the new image has a smaller ip, it is chosen as the best image of the 
3D grid point. Otherwise, the old one is kept as the image of the given 3D grid point. This process is repeated for all images at the 
station. In the end, the image with the minimum ip is defined as the image for the grid point. 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B5. Amsterdam 2000. 183