What features are extracted from the objects in the VBAI Classification?

[kdalton]

I am trying to use the Classification option in VBAI, but can't find the actual features anywhere, nor any discussion of the details of the extracted features.  I need know what features are used in the feature vector to best process/enhance my images to optimize these features.  How can I see what VBAI uses for the feature vector?  It's not important to see the actual numbers, I'd like to know the names (and algorithms would be useful too).  BTW I am a VBAI user,  and do not have LabView.

Thanks!

K

[gregstoll]

kdalton -

A description of Classification is available in the NI Vision Concepts Manual.  The features used for Binary Particle Classification are described here.  Hopefully the descriptions will be informative.

Greg Stoll
IMAQ R&D
National Instruments

[a.sardella]

Dear Mr Stoll,

could you provide me with a more detailed description of the 7 feature used by the LV particle calssifier. I would like to write a custom one based on these feature but extending it also to some gray level feature like mean gray level of particle and the histogram spread.

Best regards

A. Sardella

Italy

a.sardella@brevetti-cea.com

[DJ_L.]

Hey A. Sardella,

I was wondering what kind of information that you are looking for when you ask about the "7 feature" used by the LV particle calssifier. What kind of a detailed description are you looking for? How are you wanting to write a custom "feature 7"? What are you wanting to do overall? The more information that you can provide with what you are wanting, the more we can help you out. Thanks.

Regards,

DJ L.

[a.sardella]

Dear DJ L in the Vision Concepts Manual the binary particle classifier use the following features:

• Feature 1 describes the circularity of the sample.
• Feature 2 describes the degree of elongation of the sample.
• Feature 3 represents the convexity of the sample shape.
• Feature 4 is a more detailed description of the convexity of a sample
• shape.
• Feature 5 is used for the discrimination of samples with holes.
• Feature 6 is used for more detailed discrimination of samples with
• holes.
• Feature 7 represents the spread of the sample.
• Feature 8 represents the slenderness of the sample.

I would like, if possible, to have a more detailed description of the feature used in terms of the NI Particle Measurements where used and eventually the name of the alternative algorithm used (combined hu moments, fourier desciptor.....)

Best regards

A. Sardella 

[DJ_L.]

Hey A. Sardella,

Unfortunately, the Vision Concepts Manual is the only information available to you about the classification algorithms that we use. Other than that, we would have to look at the code of the algorithm that we use to give you a more detailed description, but we cannot release that information. Sorry for any inconvenience that this may cause, but you will have to work with the information that the Concepts Manual provides. Good luck in your application.

Regards,
DJ L. 

[nesar_m]

I am working on a project and I used k-NN classifier but I dont know what features should be used and what is the feature extraction approach in classification by labview. I cant find Ni vision concepts manual pdf. can you email this for me?

best regards

N_mohamadi85@yahoo.com

[Brad]

The concepts manual is now a chm file located here on my machine: 

C:\Program Files (x86)\National Instruments\Vision\Help\NIVisionConcepts.chm

The Binary Particle Classification section should be helpful (use the search tab to find it).

[nesar_m]

Thank you dear Brad. I saw that. Are the feature extraction approach and also the appropriate features determined by the software itself? cant The person participate in select them?

[Brad]

The built-in classifier cannot be modified, but if you want to have control over the parameters used for classification, you can use a custom classifier where you can determine your own criteria and pass in an array of doubles that represent each object (typically results from particle analysis). Just make sure the array is the same size for all the objects and the custom classifier will use the same engine to return the closest class name based on your custom feature vector.

Check out C:\Program Files (x86)\National Instruments\Vision Assistant\Solutions\Manufacturing\Legos\Custom Classifier.llb\Custom Classifier.vi to see how the custom classifier array of data is created so the engine can return the closest matching class with similar array of data.

Hope this helps,

Brad