Abstract:
Precast concrete components are complete structures of reinforced con-
crete which are pre-manufactured in a factory and then delivered to
the construction site ready-to use. These components normally ar-
rive just in time and are directly incorporated into the construction
on arrival. Because of that, falsely labeled components or deviations
from the speci cation are very harmful as they are often recognized too
late. To cope with this, the components are automatically detected in
camera images and veri ed to 3D model speci cations of the required
parts. To do so, distinctive parts or shapes of the detected component
have to be matched between the image to the model to compute a 2D-
3D mapping. Afterwards, the shape and structure can be compared
and veri ed. Multiple images of di erent views are necessary to com-
pletely verify the structures. For this task, suitable feature detection
algorithms is determined and implemented to compute 2D-3D corre-
spondences. Afterwards, the pose of the camera is estimated and 3D
points are projected to 2D. Finally, the shapes of the components in
the 2D image and the 3D model is compared and veri ed; di erences
are visualized.
