Prev: JESUS in the QURAN
Next: how to obtain frequency and power/frequency vectors using Hpsd(...) function
From: Dana on 25 Apr 2010 13:25 hi everybody I have a project in pattern recognition, I'm trying to make a Banknote classifier software, but I have troubles in features extraction. I thought of color as the main feature, but the results for banknotes samples were not acceptable. the samples are taken by digital camera (not by scanner). have anybody done anything similar?, or any enhanced color recognition code?, or any ideas? thanks in advanced
From: Thomas Britton on 25 Apr 2010 14:49 "Dana " <areejon(a)hotmail.com> wrote in message <hr1tu3$cqn$1(a)fred.mathworks.com>... > hi everybody > > I have a project in pattern recognition, I'm trying to make a Banknote classifier software, but I have troubles in features extraction. > > I thought of color as the main feature, but the results for banknotes samples were not acceptable. > the samples are taken by digital camera (not by scanner). > have anybody done anything similar?, > or any enhanced color recognition code?, > or any ideas? > > thanks in advanced Do you have some example images?
From: ImageAnalyst on 25 Apr 2010 15:23
Color could be a main feature but it's not sufficient. I wouldn't pass your project if all you did was to find the mean RGB of the image and say that it matches that of a bank note. I'd want you to put in several other things, such as checking the variance also, plus some other features like the area fraction of some particular color, PSNR compared to a known, perfect note, or something else. Otherwise it's too trivial of a student project. Even with checking mean, variance, area fraction, and PSNR it could all be done in a few hours. If it's a semester long project you'll have to put more effort into it than a few hours. If it's homework due next week then those may be sufficient. |