International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004 
The second series of experiments involved the scanning of five 
targets that were mounted on a wall, from different angles and 
distances. The reference dataset is created using four merged 
scans for each one of the targets, collected from a 5m distance 
with the scanner facing the wall. The other datasets were 
acquired by scanning the same targets from distances of 3 and 
10 meters and with the z-axis of the scanner's system forming a 
90° a 45° and a 15° angle. This way, 6 datasets were created 
and the results derived following the same procedure as 
previously are summarized in Table 6. 
The results of the ‘maxrad’ and ‘maxrad4’ are very unstable, 
ranging from 7mm to more than 25mm. As for the results of the 
*radcent! method, the mean absolute error is within the 
accuracy specifications of + 6 mm of the Cyrax system. The 
results of the ‘fuzzypos’ and ‘fuzzyposfine’ methods are once 
again the best, especially those of the ‘fuzzyposfine’ method 
(i.e. 0.4mm for every case at the 10m distance). For the case of 
the 3m distance, it seems that as the angle becomes smaller, the 
Mean Absolute Error tends to be greater. The performance of 
the *fuzzyposfine' method appears to be even better for the 10m 
distance, in comparison to the case of the 3m distance. As for 
the other methods, the results of the 'gridrad' and 'delrad" 
algorithms are similar to those of the ‘radcent’ algorithm. For 
the *fuzzygridrad' and *fuzzydelrad' algorithms, the results are 
also quite satisfactory, with a Mean Absolute Error of about 
2mm, which is also better than the accuracy specifications of 
the system. 
5. CONCLUSIONS 
[n this paper, the repeatability of measurements obtained by the 
Cyrax 2500 laser scanner, a widely used system, has been 
examined. It was shown that the repeatability is high for all 
datasets collected in laboratory conditions. Furthermore, the 
properties of the Cyrax reflective targets were thoroughly 
examined and presented. 
A number of new algorithms for target identification have been 
proposed. With datasets collected in different experiments, both 
the internal and external accuracy of all of the algorithms was 
examined. It was shown that using Fuzzy clustering techniques 
gives insight to the processing of the data of the targets, and 
therefore it is highly recommended as a tool for further 
research. The results of the proposed methods, especially those 
of the ‘fuzzypos’ and ‘fuzzyposfine’ methods, are proved to be 
very accurate and reliable. These methods can be used when 
there is a high demand in accuracy, for instance for 
metrological experiments, deformation monitoring, registering 
multiple scans etc. They may be more demanding in 
calculations compared to other methods, but the results are 
substantially better. 
For future work, more experiments need to be conducted in 
order to evaluate the performance of the algorithms for various 
resolutions, greater distances and non-laboratory conditions. 
6. ACKNOWLEDGMENTS 
This work and the first author are supported by a research grant 
(Program Thalis) from the National Technical University of 
Athens. The help of Hristos Gounaris, undergraduate student of 
NTUA, in conducting a number of experiments is also 
acknowledged. 
7. REFERENCES 
Balzani M., Pelegrinelly A., Perfetti N., Uccelli F., 2001.A 
Terrestrial 3D Laser Scanner: Accuracy Tests, Proceedings of 
18^ International Symposium CIPA, Potsdam, Germany, 
pp.445-453 
Bezdek, J.C., 1981. Pattern Recognition with Fuzzy Objective 
Function Algorithms, Plenum Press, New York 
Lichti D.D., Stewart M. P.M., Tsakiri M., Snow A.J., 2000. 
Benchmark Tests on a Three-dimensional Laser Scanning 
System, Geomatics Research Australasia, vol. 72, pp. 1-23 
Gordon S., Lichti D.D., Stewart M. P.M., Tsakiri M., 2001. 
Metric Performance of a High-resolution Laser Scanner, 
Videometrics and Optical Methods for 3D Shape Measurement, 
pp. 174-184 
     
  
  
  
   
  
  
  
   
  
  
  
   
  
  
  
   
  
   
  
  
  
   
   
  
  
   
   
  
  
  
   
  
  
  
  
  
   
   
  
  
  
  
  
  
  
   
     
   
   
    
   
    
  
  
   
  
  
  
   
   
  
   
   
Sectio 
KEY Wi 
Modellir 
ABSTR, 
We pres 
achieving 
rapid anc 
they hav 
laser scat 
superiori 
visible in 
of laser 
strategy 
represent 
main dat 
planes ar 
CAD mc 
image gr 
model m: 
similar tc 
all imag: 
available 
The resu 
and imag 
As built 
purposes 
Moreovei 
Augment 
and oper 
of these t 
environm 
alignmen 
the Secti 
Delft has 
Reality. f 
users (S' 
everythin 
more flex 
objects a 
services « 
3D mode 
for align 
critical. T 
* 
Corres