Hardware Reference
In-Depth Information
Figure 2.7
Industrial application that requires a visual recognition of objects on a con-
veyor belt.
Two tasks (one for each camera) dedicated to image acquisition, whose objec-
tive is to transfer the image from the camera to the processor memory (they are
identified by
acq1
and
acq2
);
Two tasks (one for each camera) dedicated to low-level image processing (typical
operations performed at this level include digital filtering for noise reduction and
edge detection; we identify these tasks as
edge1
and
edge2
);
A task for extracting two-dimensional features from the object contours (it is
referred as
shape
);
A task for computing the pixel disparities from the two images (it is referred as
disp
);
A task for determining the object height from the results achieved by the
disp
task (it is referred as
H
);
A task performing the final recognition (this task integrates the geometrical fea-
tures of the object contour with the height information and tries to match these
data with those stored in the data base; it is referred as
rec
).
From the logic relations existing among the computations, it is easy to see that tasks
acq1
and
acq2
can be executed in parallel before any other activity. Tasks
edge1
and
edge2
can also be executed in parallel, but each task cannot start before the associated
Search WWH ::
Custom Search