An automatic procedure that is able to estimate the orientation 
parameters of a digital image is described. 
All the information that is needed is extracted automatically 
from the laser scanner DTM and from the digital images that 
have to be oriented. 
2. AUTOMATIC ORIENTATION OF DIGITAL IMAGES 
The proposed procedure requires the presence of some 
reflecting targets on the scanned object. This targets realise the 
control point set that is necessary for the estimation of the 
orientation parameters of the same object digital image. 
A specific software package has been implemented in order to 
perform the following phases: 
* automatic extraction of the reflecting target 
coordinates from the DTM acquired using a laser 
scanner (point set 1); 
* automatic extraction of the reflecting target 
perspective coordinates from the digital image, by 
means of a conventional least squares matching 
algorithm (point set 2); 
® correlation of point sets 1 and 2 to associate, the 
correspondent image point (the result of this last step 
creates the necessary input data for the estimation of 
the orientation parameters) to each target, extracted 
from the DTM; 
* orientation or calibration of the digital images by 
means of a traditional bundle solution. 
The software has been implemented on a standard PC and has 
been developed using Visual Basic language. 
The following paragraphs describe the used algorithms in detail 
and a practical example demonstrates their performances. 
2.1 Placement of the reflecting targets 
Before the laser scanner and digital image acquisition phases 
are carried out, a sufficient number of reflecting targets have to 
be placed on the object. 
  
Figure 2. Example of reflecting targets placed on the Ramses II 
statue (Egyptian Museum — Turin) 
The location of the targets must be carefully planned in order to 
guarantee the presence of a sufficient number of them inside the 
DTM and the image that has to be oriented (e.g. at least 4 points 
if a metric camera is used and at least 7 points if the image has 
been recorded using a non metric camera). 
The relative positions of the target have to fulfil the well known 
requirements for a correct estimation of the orientation 
parameters (e.g they have to circumscribe the object). 
The size of the reflecting targets have to be chosen in order to 
guarantee the automatic extraction from both the DTM and the 
digital image. 
If in both cases the targets size is directly dependent on the 
taking distance, some more constraints must be taken into 
account. 
For a correct laser scanner acquisition the targets have to be at 
least two times larger than the laser instrument scan rate (e.g. 
using the Riegl LMS Z210 laser scanner at the scan rate of 
0.040 gon, the target can be a circle with a diameter of 35 mm if 
a taking distance of 50 m is adopted). 
As far as digital image is concerned, the focal distance of the 
used lens and metric resolution of the digital image also have to 
be considered. 
Using a focal length of 40 mm, a taking distance of 40 m and a 
pixel dimension of 10 um (a good resolution for terrestrial 
images), the image with a diameter of 35 mm is of about 35 um 
(almost 3 pixels); in this case traditional matching algorithms 
can run with good results. 
2.2 Reflecting target coordinate extraction from laser 
scanner DTM 
The aim of the first step is to extract the reflecting target 
coordinates from the DTM produced by the laser scanner. 
The user extracts the portion of the DTM that is useful to 
describe the shape of the scene recorded by the digital image 
that has to be oriented. 
  
Figure 3. Digital image and extracted DTM 
This operation can be easily done using the acquisition software 
of the instruments; the coordinates of the selected points are 
exported into ASCII format. 
Some initial parameters have to be fixed by the operator to 
allow the automatic extraction of the reflecting targets. 
The user defines the size of the used signals and the minimum 
value of their reflectance. 
—322—