Logo inspection using an SVM approach.

[SimonAldred]

I am interested in applying the SVM algorithm to the problem of logo inspection in order to identify defective logos. Does anyone have any ideas where to start?

[CarlFHellwig]

Hi SimonAldred!

There are functions for LabVIEW in the Vision Development Module that enables you to work with SVM: http://www.ni.com/labview/vision/

With the module comes also a few classifying example codes that can help get started. The NI Vision Concepts Help can also be of assistance with the theory of VSM.

How familiar are you with the theory behind the SVM algorithms? Have you tried creating an algorithm in LabVIEW, or are you just looking for pointers?

Here are also a few resources I hope can be of help:

https://decibel.ni.com/content/groups/london-labview-user-group/blog/2014/01/13/multiclass-classific...

https://lavag.org/topic/13082-support-vector-machine-classifier/

http://zone.ni.com/reference/en-XX/help/370281P-01/imaqvision/imaq_train_svm/

[SimonAldred]

Hi Carl, I didn't realise that vision had some svm code. I know it has pattern matching and golden template matching, both of which seem pretty straightforward to use. I did download a library that contained one simple SVM example, nothing to do with vision, so I was wondering how to extend this to a vision problem. I have taken some tentative next steps by extending this program to work with vectors of more than 2 dimensions so this is some kind of progress. I am familiar with the SVM algorithm, just not so clear how to apply it to something like logo inspection. In particular what my input parameters should be. Anyway thanks for the pointer towards vision, I will have a closer look. Simon