The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008 
1184 
of 3 comparing digital image matching in DMC images to 
manual point identification in analogue images. 
Analog semiaut. 
DMC semiaut. 
DMC MatchAT 
pm 
pix. 
pm 
pix. 
pm 
pix. 
x 4.83 
0.32 
2.85 
0.24 
1.23 
0.10 
y 4.27 
0.29 
2.35 
0.20 
1.12 
0.09 
Table 1: Photogrammetric residuals in pm and pixel (Alamus et 
al., 2005) 
Image pointing accuracy using AAT techniques with analogue 
images is in the level of 4 to 5pm (i.e. l/3 rd of a pixel when 
scanned at 15pm), with DMC images it is in the level of 1.1 to 
1.4pm (i.e. 1/10 th of a pixel). 
2.2 Small scale flights 
To test the performance of aerial triangulation check points 
from several blocks have been analyzed. The analysis has been 
carried out in two different data sets: blocks with a GSD of 
45cm, studied in this section and blocks with a GSD of 7.5cm, 
studied in the next section. 
ratio” (when taking b as the base between images from different 
strips) plays an important role (and that value is similar in the 
analog and digital scenarios). 
Figure. 1 : Aerotriangulation check point accuracy vs. year of 
flight 
From year 2004 to 2007 ICC, in collaboration with the PNOA 
project, is flying and aerotriangulating half Catalonia at a GSD 
of 45cm, which corresponds to approximately 5000 DMC 
images every year. The check points used in the comparison 
were obtained from the 50 blocks used to aerotriangulate the 
20 000 above mentioned images. It is important to point out that 
the used blocks are taken as they are without any kind of 
additional optimization. Therefore, the accuracies shown in the 
paper corresponds to a day to day work in a production 
environment rather than the result of an academic effort. 
The accuracies of the check points shown in table 2 
demonstrate that there is a slight improvement in the three 
coordinates when the DMC camera is used. 
RMS X 
(m) 
RMS Y 
(m) 
RMS H 
(m) 
N.Checks 
Analog cameras 
0.22 
0.20 
0.28 
90 
Digital (DMC) 
0.21 
0.19 
0.26 
280 
Table 2: Comparison of the analog vs. digital check point 
accuracies for 45cm GSD 
In 2004 the ICC already had many years of experience in aerial 
triangulation of analog projects, but with the reception of the 
first DMC, in late 2004, some parts of the workflow had to be 
readjusted and some “know-how” had to be acquired (use of 
self calibration parameters, quality of GPS/IMU information, 
error propagation...) During the first 3 years of DMC operation 
an improvement on the quality of the aerial triangulations has 
been observed, which is attributed to the learning curve of the 
new technology. Figure 1 shows the check point accuracy by 
year and a slight yearly improvement can easily be observed. 
The matching accuracy improvement observed in the 
correlation of digital images has not been translated to a big 
improvement in the horizontal accuracy because the check 
points are measured manually without any automatic or 
semiautomatic support. Concerning the small improvement in 
the vertical component, despite of the worse b/h factor, it has to 
be noticed that most of the check points are measured in images 
from different strips (flown at 25% side lap) so the “side lap b/h 
2.3 Large scale flights 
Regarding the results from large scale flights, the check point 
accuracies from 25 different blocks are shown in table 3. The 
data corresponds to production urban projects with a GSD of 
7.5cm for both digital and analog (scanned at 15pm) flights. 
RMS X 
(m) 
RMS Y 
(m) 
RMS H 
(m) 
N.Checks 
Analog cameras 
0.032 
0.050 
0.059 
19 
Digital (DMC) 
0.035 
0.041 
0.058 
117 
Table 3: Comparison of the analog vs. digital check point 
accuracies for 7.5cm GSD 
For the combined horizontal accuracy, a very slight 
improvement in the digital flights can be observed while the 
vertical component of the accuracy remains stable. It has to be 
pointed out that the relative accuracy of the check points is at 
the level of 2cm, therefore it is possible that a portion of the 
residuals from the check points is due to the uncertainty of the 
ground coordinates. 
As in the previous section the accuracy of the three components 
is very similar in analog and digital blocks. The explanation 
might be the same as in the previous section. The image 
observations of the check points are given by manual 
measurements (not taking advantage of the superior image 
correlation of the digital image) and in different strips (taking 
advantage of the side lap b/h ratio). 
3. DEM GENERATION 
This section describes and discusses the obtained accuracies of 
automatically DEMs derived for the same area from three 
different flight configurations The DEM are created using the 
commercial software package Match-T of the Inpho Company. 
The 3 flights were taken in a 3 year time difference with an 
analog RC30 camera at a ground pixel size of 45cm and with 
the DMC at a GSD of 45cm and 20cm. The major 
characteristics of the 3 corresponding blocks are listed in table 4.