The process of recognition is based on comparison of two shapes. An unknown shape must be compared to a set of known shapes in a database, and determine how similar they each is. Arkin et al [3] proposed a polygonal contour similarity measure that is:
- a metric;
- invariant under translation, rotation, and change of scale;
- easy to compute.
Basically we take a polygon such as:
And represent it in a much more useful form:
If you want to see the math behind this, and how the comparison process actually works,
