Prakt. Met. Sonderband 38 (2006) 133 
The Figure 6: Object 2: Images of the two surface regions with manually segmented profile 
paths. Left: surface S1. Right: surface S2 
Length (mm) Length (mm) 
Figure 7: (a) Corresponding fracture surface profiles of both halves that were manually 
a extracted. Solid profile: surface S1, dashed profile: surface S2. (b) Corresponding surface 
> (object profiles after the automatic registration. The dashed profile on S2 corresponds to the 
esponds marked profile in Figure 7a, the corresponding solid profile on S1 has been automatically 
ans with extracted by the registration algorithm. 
After the It should be noted that both in micro-ductile and brittle fracture often secondary cracks 
as been appear, leading to rotations and lateral shifts of some fracture surface regions. This makes 
analyzed the measurement of COD; and the automatic registration of the two fracture surfaces much 
I. 6. The more difficult. To further improve the procedure, a possible solution would be to apply a 
ye the fragmentation of the 3D model, especially for regions where high misfits occur, and 
utomatic perform the registration in parts. This should work for brittle fracture; for ductile fracture it 
SOIC AG will be probably very difficult to distinguish between misfit caused by plastic deformation 
4 and misfit caused by secondary cracks. 
5 able to 
the cast 
ct which 6. CONCLUSIONS 
d height 
nall void An algorithm has been developed that automatically performs the task of registrating two 
~ surface corresponding fracture surfaces together. The algorithm uses 3D models of the surfaces 
nisfit are that have been generated by the software package MeX from stereoscopic image pairs or 
he misfit triplets. The registration algorithm is able to process fracture surfaces with large 
neral the deformations as long as there are significant parts in the 3D model that can be modelled 
MC. by the same Euclidean transformation. Experiments have shown that the algorithms