Its a good idea and Im not saying it wount work or it shouldn't be attempted, but flowers do not distinguish different cultivars alone. Many are very similar, especially if they are in the same subgroup. The reason they could be different could be because of size (as an example) and flowers would have nothing to do with it. If you would like to create a database, I would reccomend using traits that can be distinguished easily (this may include flowers but does not have to limited to them, foliage can also help identify plants).
Professionally, bananas are characterized by a big list of different things, because flowers alone cannot always tell what cultivar it is.
http://bananas.bioversityinternation...26/53/lang,en/
Bioversity International is constantly working towards improving thier MGIS (Musa Germplasm Information System), they already have passport data with picrures on many different cultivars and are working towards getting pictures and descriptors assigned to each of thier accessions.
Also the Musalogues from Bioversity International (along with the other germplasm catalogs,
http://bananas.bioversityinternation...25/54/lang,en/) are wonderful resources that you should look into.