8) 
se 
ve 
+1, 
the 
0) 
1) 
2) 
1 Or 
13) 
Aluir Dal Poz 
e Dy. isthe predicted residual of b or b* and is the second element of v; 
e Sa" \s the standard deviation of n... and is the square root of the first element of diagonal of c; 
® Spy is the standard deviation of nn, ,. and is the square root of the second element of diagonal of c; and 
e — Wy,. and. Wy are the normalized corrections. 
The observations a (or a*) and b (or b*) do not have errors when the values w,,« and Wins, With the significance level a , 
are contained in the intervals: 
-Na p < Wa «Nan» and -Na D € Ww «Na n (14) 
where, Na / is extracted from normal curve. 
3 RESULTS WITH REAL DATA 
The entire system has been implemented in C language, with a command line user interface and it has been designed to 
be automatic. All the routines including estimation and image processing were in-house developed. 
    
3 : F ka ; Ri " 
A EN M 3 n S R 
i. lui Od ig 
Figure 4. Results from the pipeline for feature extraction and matching 
(a) Window automatically extracted from the original image (grouping 3 — highlighted in 
table 1); 
(b) Edges obtained by applying the Nevatia and Babu filters and automatic thresholding; 
(c) Thinned edges with non maxima suppression and filtering of isolated elements; 
(d) Straight lines after line fitting. The highlighted lines (1) and (2) were previously 
known in the object-space and the line (2) was correctly matched. 
Experiments with real data were performed in order to verify the potential of the proposed method. A scanned aerial 
photograph (1:8.000) of Rio de Janeiro City was used. The original image was resampled to a pixel size of 75m in order to 
facilitate the debugging of the C codes and to avoid a large data set. Experiments with the original data are being conducted 
and will not be reported in this paper. 
  
  
  
Selected Number of straight lines 
Groupings Object-space Image-space 
l 4 6 
2 6 10 
3 2 6 
4 2 9 
5 2 4 
6 2 10 
  
Table 1. Image and object-space groupings 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 211