Partical count

[LV_Outsourc]

Hi

Need to count the number of parts (see the picture).
Tried the following example vi:

Pattern Matching.vi

ColorPatterMatching Example.vi

Geometric Matching Example.vi

Particle Analysis Example.vi

Not one of the methods did not give good results.

Advise method of implementation please!

[MoviJOHN]

1.  You need more pixels.

2.  You need good focus.

3.  You need uniform illumination.

The more pixels you can throw at an image, the better the classifiers, and other analysis functions will work.

Unless you have distinct color feature that are helpful in identifying individual parts, a grey scale camera will give better results.

You won't get very far with room lighting and a consumer grade USB camera.  You might also consider that the sample needs an agitator to separate out the parts, for better results.

[LV_Outsourc]

Maybe someone saw on the Internet a good description of the method of separation of particles that suited to my case?

[Emily-C]

Hi LV_Outsourc,

MoviJOHN is right that in order for this to work well, you'll need better images to start with. I played around with this a bit, and was able to get some particle analysis working for the SH1 image, but when you're using particle analysis you won't be able to recognize overlapping objects. All I used was a thresshold, then particle analysis. With threshhold values of 0 and 181, I was able to count 49 objects (screenshot attached showing the results). You might be able to handle slight overlapping by using matching instead (pattern or geometric) but I don't think that will work very well unless you can get a clearer image.

For SH2, you really would need to improve your images. With the low resolution, inconsistant lighting, overlapped objects, and looking at the objects from different angles, there just won't be enough consistency to use particle analysis or pattern matching. 

I hope this helps! Good luck with your application.

[Hornless.Rhino]

You can also look at using circle detection, but I guess that's not strictly using the particle analysis stuff.