f them inside the 
. at least 4 points 
if the image has 
1 the well known 
the orientation 
bject). 
losen in order to 
he DTM and the 
ependent on the 
t be taken into 
ets have to be at 
t scan rate (e.g. 
the scan rate of 
eter of 35 mm if 
| distance of the 
age also have to 
e of 40 m and a 
n for terrestrial 
of about 35 uum 
hing algorithms 
mn from laser 
eflecting target 
' Scanner. 
iat is useful to 
le digital image 
TM 
sition software 
ted points are 
he operator to 
rgets. 
| the minimum 
  
This value depends on the taking distance of the laser scanner 
and must be checked each time on the acquired data. 
The reflecting targets always show an image that is larger than 
their real size; this is due to the high reflectivity of the signals 
compared to the reflectivity of the surrounding materials (light 
diffusion). For this reasons the software asks for the size of the 
apparent images of the targets. 
Finally the operator indicates the linear and angular tolerances 
and the beam divergence of the used instruments. 
The software sequentially reads all the recorded reflective 
values and stores the selected point coordinates in a separate 
matrix (the ones with a higher reflective value than the 
minimum value chosen by the operator at the beginning). 
  
  
Impostazione caratteristiche dello s Fup 
x 
  
  
Markers utilizzati 1 Precisione dello strumento T Tollerenze di contranto 
Dimensione dei markers utilizzati per ia. d 
preseanalizzazione dei punt us : 
Diemetro dei mark LS em 
Valore minimo di rifletiività da utilizzare 180. [8/258] 
  
  
  
  
  
Ame cono 
iv. imposta la tolleranza lineare di confronto come [4 volte il diametro dei markers 
| Impostalatollerenza angolare di conitonto come [a volte illa precisione angolare 
  
  
  
  
  
| Annulla | Applica 
  
  
Figure 4. Initial parameter input window 
If the adopted criteria for the definition of the target size is 
considered, it appears clear that more than one of the selected 
points can belong to the same target. 
In order to group together all the points that define a single 
reflecting target, the software performs the following steps: 
e the first of the selected points is considered as a point 
of the target to be searched for (origin); 
e the distances of the remaining selected points from the 
origin are calculated; 
e all the points which have a distance shorter than the 
size of the apparent image of the target (previously 
defined by the operator) are considered to belong to 
the same marker and extracted from the matrix; 
e the origin point and all the extracted points in the 
previous step are then recorded in a separate sub- 
matrix. 
This procedure is iterated until the high reflectivity point matrix 
is empty. 
The sub-matrices that arise from the described procedure 
contain all the points acquired by the laser scanner on the same 
target. 
In order to avoid outlier and noise acquisitions, the target 
coordinate computation is conceived as a robust estimation. 
This operation can be summarised as follow: 
e the coordinates of the target are calculated using the 
median of the selected coordinates; 
e all the points which have a higher distance than the 
linear tolerance of the used instruments from the 
original point are considered outliers and rejected; 
e the means of the remaining point coordinates of the 
are considered to be the correct coordinates of the 
target. 
82 high reflectivity points have been selected using the DTM of 
figure 2. 
The grouping procedures have defined 13 reflecting targets. The 
filtering procedure gives the final coordinates of the reflecting 
targets. 
In order to test the implemented algorithm, the same points have 
been surveyed by a total station. 
Table 5 shows the statistics of the differences between the 
relative distances of the targets calculated form the two sets of 
coordinates (one estimated by the software and the other 
determined by the total station). The target n# 20 has been used 
as the origin of the compared distances. 
  
  
  
  
  
  
  
Target Distance 20-# Discrepancies 
# [mm] [mm] 
| Software | Total Station | : 
20 0 0 0 
21 426 423 3 
22 1006 1004 2 
24 1816 1812 4 
25 1987 1992 -5 
60 90 89 1 
61 464 466 -2 
62 605 610 -6 
63 791 796 -5 
65 1173 1166 6 
66 1246 1246 
67 1376 1380 -5 
68 1649 1641 8 
Table 5 
The used DTM has been acquired using the Riegl LMS 25HA 
laser scanner which measures distances with a standard 
deviations of 8 mm. 
The results confirm the correct and complete automatic 
extraction of the reflecting target from the laser scanner DTM. 
The set of the reflecting targets coordinates extracted from the 
DTM is named “point set 1” in the following. 
2.3 Reflecting target extraction from digital images 
The second step of the procedure computes the coordinates of 
the reflecting target perspective images recorded on the digital 
image that has to be oriented. 
A traditional least square matching procedure is used. The target 
matrix represent the true shape of the signals, scaled according 
to the used focal distance and the approximate taking distance. 
It is well known that to speed up the matching of pixel 
approximation, a approximate position of the searched point has 
to be defined. 
The software adopts an automatic procedure. Approximate 
exterior (e.g. coordinates of the projection point C and the o, ¢ 
and x angles) and inner (e.g. focal distance) orientation 
parameters of the image are usually known. 
These approximate values allow one to define an approximate 
plane of the image where it is possible to project the coordinates 
of the "point set 1" as defined in the previous paragraph (see 
fig. 6) using collinearity equations. 
The search matrices are then centred on these points and the 
size has to be large enough to avoid any differences between the 
true image (the one that has to be oriented) and the virtual 
image (the one created by the software). 
The reflecting targets are not well contrasted inside a digital 
image acquired in the visible field. 
—323—