aA
Oy NO NS
[4 NU AEN
NS NN
© OO
Picture 9: Verification results of arcs varying in size ( 85
to 30 pixel with 360 degree )and angle (22,5, 45, ... 315
degree).
Further 2D primitives within the DXF-Format may be
projected to one of the presented types. Composed
structures like squares, rectangle, etc. may be defined
for verification purposes. Complex structures are
verified by a sequence of primitive verifications (picture
11).
The application of the described proving operations is
not limited to two dimensional objects. Once the 3D
model primitives have been projected to the image
frame using a more powerful transformation they may
be verified in the same manner. Some important CAD-
models like Boundary Representations (BReps) are
based on the primitives we are using here and we hope
to verify these pretentious models after some further
work.
| verify line | | verify circle | | verify arc | | verify ... |
Picture 10: Hierarchy of the discussed verification
routines
8. THE ADVANTAGE OF PROVING OPERATIONS
The proving operations described above consist of a
® type and parameter specific pixel segmentation
* type specific approximation
* similarity function
The combination of these tasks which are usually
spread about different levels like segmentation, feature
extraction and classification within a single operation is
the foundation of the obtained modularity.
Picture 11: Intensity and direction image of a tin product and those elements, verified by the described algorithms.
The discussed proving operations offer the capabilities
to decide wether there is a special image portion.
Because of the modular concept in using primitives,
these special portions cover single primitives,
substructures as well as complex structures, including
such useful basic forms like squares, rectangles and
others. If there is a manageable set of hypothesis this
approach prevents a costly bottom up check with
unknown borders, types, orientation and location.
Within an interactive image analysis system such
verification routines may become a powerful tool for
recognition and conversion tasks. But even when there
is no user to secure the number of hypothesis usually
there is enough a priori knowledge useful for
limitations. Therefore we are engaged in developing a
control structure suitable in the depicted context.
The application of the described proving operations is
not limited to two dimensional objects. Once the 3D-
model primitives have been projected to the image
frame using a more powerful transformation they may
be verified in the same manner. Some important CAD-
models like Boundary Representations (BReps) are
based on the primitives we are using here and we hope
to verify these pretentious models after some further
work.
9. CONCLUSIONS
We have discussed algorithms for the verification of
graphical primitives. The operations are organized to
work in a strong model driven interpretation process.
The gradient direction image we are working at is much
closer to the geometric CAD-models than a captured
greyscale image and is close enough to the captured
data to prevent the influence of numerous thresholds.
The described operations work on 2D primitives in 2D
images, nevertheless they may be used in 3D context
after a suitable projection.
