We present a model-based vision system to recognize planar objects which can be
represented by a set of lines. Our system uses the projective invariants of five lines.
However, a set of five lines can produce up to 120 different invariant values. As the
number of models in a model database becomes larger, the size of search space to
find a corresponding model in the database may increase exponentially. In order to
solve this problems, we introduce a line convex hull(LCH) and an indexing logic
filter(ILF). The line convex hull classifies a set of five lines into one of twenty-one
different types of convex hulls and also provides an unique ordering to five lines.
Using the types of line convex hulls, ILF computes an integer indexing value to
represent a set of six lines.
By combining the LCH and the ILF, the proposed scheme greatly improves
search speed to find candidate model features in the model database that match
scene features. We have performed a series of experiments on real images of the
fifteen synthesized models of welding panels. The system has successfully
recognized models in the database as well as the corresponding features.
represented by a set of lines. Our system uses the projective invariants of five lines.
However, a set of five lines can produce up to 120 different invariant values. As the
number of models in a model database becomes larger, the size of search space to
find a corresponding model in the database may increase exponentially. In order to
solve this problems, we introduce a line convex hull(LCH) and an indexing logic
filter(ILF). The line convex hull classifies a set of five lines into one of twenty-one
different types of convex hulls and also provides an unique ordering to five lines.
Using the types of line convex hulls, ILF computes an integer indexing value to
represent a set of six lines.
By combining the LCH and the ILF, the proposed scheme greatly improves
search speed to find candidate model features in the model database that match
scene features. We have performed a series of experiments on real images of the
fifteen synthesized models of welding panels. The system has successfully
recognized models in the database as well as the corresponding features.