This project is a recent idea of mine and a friend to compare keys. The problem is that a locksmith needs to cut blanks for the keys which his customers hand him. However, these keys are all different, and he spends a substantial amount of time comparing these keys to those on his wall to find the correct match. Using this program (which is not yet completely implemented), he can use a photo of the key to compare to other keys in a database, producing similarity scores of which 0 indicates the greatest similarity. The actual calculations involve calculating a "shape context" (read more at http://en.wikipedia.org/wiki/Shape_context) based on several points around the edge of the key, and then creating histograms of this data. These histograms are compared with the chi-squared test. Several components have yet to be implemented, and issues yet resolved, such as: the fact that there should involve a thin plate spline transformation in the comparison which I cannot implement because I do not yet understand the advanced integral notation (http://en.wikipedia.org/wiki/Thin_plate_spline); the fact that holding more than a few keys in the database would involve an excessive amount of data; getting a perfect key image is unfeasible in most situations.

All work (c) Eric Salina 2014-2015 except where noted otherwise (as in the Hungarian Method implementation).