Thread Subject: color segmentation

hi everybody

i have a project for my pattern recognition course, and planning on
designing a system the recognizes defective bottles on a conveyor belt
with a low hue color background (Red).

based on:
1) white spots on the blue bottle (means that the bottle is defective)
2) bottle neck
3) bottle base color (white -> defective)

so please what do you guys think about this project

and also, what methods (classifiers,...etc) that could help me with
the:
- color recognition
- measuring the bottle neck

thanks in advance
Khalel Dawod

Khalel Dawod
If the presence of white is all it takes to identify defective
product, then simply look for white. You can just adapt this code I
uploaded:
http://www.mathworks.com/matlabcentral/fileexchange/26420-simplecolordetection
It's a trivial matter to change from detecting red (like I did) to
detecting white.

However, unless you have good control over the lighting, you will
likely have specular reflections (glints in the glass) that will cause
a lot of false positives. Why don't you post a photo?