T 
    
Figure 4: Visualization of tracing point vector displacements 
(magnitude x10) after 1220 s from the beginning of 
loading (test 1) 
tempo (secondi) 
n 100 200 300 400 500 800 700 200 900 1000 1100 1200 1300 
  
E 
Carico Verticale (Kg) 
8 
i 
  
  
  
Figure 5: diagram time-load for test 2, showing a sudden 
variation of loading. 
   
  
: 
  
4 
i i 
— —Ó 
| 
  
  
y 
z * 
Figure 6: Mesh representation of the tracing point positions in 
image 1 and image 11 (test 2) 
3. TRACKING THE DISPLACEMENT OF POINTS ON 
A SLIDING SURFACE 
3.1 Test goal and test setup 
In a second application, the goal was to track the movement of 
points on the superficial layer of a sand specimen, sliding along 
a sloping plane. As in the previous case, the goal was to 
compare the actual dynamic of the grains with the prediction of 
a mathematical model. This is of interest in trying to model the 
conditions that lead to landslide in this kind of terrains. The 
experiment was carried out in a channel about 4 m long and 50 
cm wide with a sand height of about 15 cm (fig. 7). The 
specimen is arranged in the sloping side of the channel and kept 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004 
in place by a gate. The specimen can be prepared with different 
sand types, with varying humidity contents, while the testing 
device can vary the inclination of the upper channel section to 
reproduce different slope angles. 
The release of the gate set the sand specimen sliding, until the 
slope change of the channel slow down and then stops the 
movement; the motion of the superficial layer of sand should be 
traced. 
  
    
  
Cameras 
  
Sand specimen 
  
  
  
  
Screw to adjust 
channel slope 
  
  
  
  
  
  
  
  
  
  
Figure 7: the testing channel and the camera's locations 
As in the previous experiment, tracking individual sand grains 
is not possible, so we resorted to spheres. Due to the test 
dynamics, we had to ensure that they were faithfully following 
the sand movement and that they could be traced against the 
sand background. To this aim, pins were fixed to the spheres 
and tests were carried out with different pin length and sphere 
size. Though this solution may not be optimal (some pin get out 
of the sand and starts rolling, becoming useless to describe the 
sand movement) visual examination of different tests confirmed 
that the sphere actually follow the sand quite closely. 
To ensure a dense description of the surface deformation, a grid 
of 12x16 spheres, spaced about 3 cm was prepared and put in 
place with drilled board, to guarantee precise positioning of 
each sphere in the channel reference system in every trial. The 
position of the targets in object space and their labels are 
therefore known as long as the gate is closed. 
Being the displacement field 3D, two synchronized digital 
cameras Basler AF 101 (1300x1000 resolution, focal length 8 
mm, pixel size 6.7 micrometers, 12 fps at full resolution) were 
employed to track the trajectories. In order to raise the frame 
rate of the camera, we took advantage of the elongated shape of 
the channel, using just half frame. We could therefore achieve 
22 fps, which proved just enough to capture the motion. 
The cameras were mounted with convergent axes over the 
channel, covering the section where the movement actually take 
place, about 1.8 m long. The camera are mounted about 1.4m 
above the specimen, with axes are convergent to the centre of 
the channel section and a base/distance ratio close to 1. With 
this arrangement, assuming a measurement accuracy of 1 pixel, 
a simulation of the spatial intersection from the cameras 
predicted an accuracy for the sphere positions of 3 mm in 
horizontal and 5 mm on Z. 
A reference frame was established on the channel by fixing 40 
targets along the top of the walls and before the gate, on the 
channel bottom. Their coordinates were determined 
photogrammetrically within a block adjustment of 8 convergent 
images taken with a Nikon D100 with a 18 mm lens, yielding 
an estimated accuracy of 0.5 mm in all directions. 
    
   
   
   
  
  
   
    
  
  
   
    
  
    
  
  
  
  
   
   
   
    
  
  
    
   
    
  
  
   
    
   
    
     
   
    
   
    
   
   
    
  
  
    
Internatio 
JC HOS 
3.2 Trac 
To track 
coordinate 
implemen 
a) 
b) 
c) 
3.3 Targ 
In each in 
Foerstner 
tracing pc 
l.s.m. usi 
candidate: 
Indeed, it 
goal. Due 
grain wert 
obviously 
targets. O 
to allow s 
spraying 
intensity. 
interest p: 
accept as 
number o 
discard th 
we had to 
due to va 
spheres. 
extraction 
test, with : 
interest pi 
sequence. 
angle, it | 
still. 
  
3.4 Labe 
To consist 
decided tc