Abstract:
Object matching has many potential applications
in industry, defense and medical science.
Most matching methods introduced in recent years
are based on the invariant representations. Main
invariants applied in computer vision are algebraic,
differential invariants and integral invariants. Our
approach in this paper uses an affine integral invariant
within a Spiral Architecture. The invariant
representation is based on the extracted object contour.
The parameter to be used for parameterizing
an object contour is derived from the enclosed area.
The Spiral Architecture posseses powerful computation
features that are pertinent to the vision process.
We present a parallel algorithm for object recognition
on clusters. Image partitioning based on Spiral
Architecture provides well-balanced load and absolutely
uniform sub-images. The cluster-based object
recognition greatly increases computation speed.
