04-14-2014 06:34 PM
04-15-2014 08:41 AM
Hello Fernan1988,
Yes, it is possible you need to have installed the NI Vision Development module:
http://sine.ni.com/nips/cds/view/p/lang/es/nid/2881
Once you install the module, please check the “OCR.lvproj” within the Find Examples section in LabVIEW.
Hope you find this information useful!
Regards,
04-23-2014 01:18 AM
I do not think NI's OCR function is very robost, it may not as good as you expected.
For example, to my knowledge (may not exactly correct), the OCR function in Labview need to segment the characters/text/number into individual one by some simple threshold method, which tend to be failed if the image background is not even enough. Anyway, if the images quality are good OCR function works well.
04-23-2014 09:49 AM
I already installed the NI Vision development Module, but I can't open the example you suggested, because it seems that I need a toolkit that I currently don't have.
Could you please tell me what to do in order to make the OCR work with different illuminations.
Thanks in advance!
04-23-2014 10:05 AM
I already saw the example, but I still can't see how this example will help me accomplish what I want. What I want is to be able to use the OCR in labview with different images in different illuminations, so that no matter the illumination, the OCR will get done correctly.
04-23-2014 05:59 PM
You could try to use the NI Vision Assistant to help you config the OCR parameters.
04-23-2014 09:42 PM
Technically speaking, what you are asking to do is impossible. Fortunately, I do imposible things all the time.
I cannot spoon feed you a solution, but I can give you a rough map to follow. First, let's set some boundries to the problem. It cannot be "any light condition". You have to address conditions of too dark (near dark), and too bright (saturation) with a simple error, and give up on those.
If you have unlimited time to accomplish the inspection (on the order of a few seconds), and if your camera supports it, you can perform an analysis of the scene at multiple exposure times, and select the exposure with the best contrast to begin you OCR.
The NI OCR stuff uses gemotric match (I think), so you will need to train each font that you want to use. Each character will need multiple templates. It is possible to build intraining into your app, so that you are prompted to fill in the blank when an unknown character is found.
You can also perform the OCR on multiple exposure images, and do dome logical analysisto determine which result is the most correct.
You can also use a dictionary list if the words will be limited in some predictable way.
If you are going to rely soley on ambient light available, I can tell you that you should give up right now. For machine vision to do its job, it needs good cameras, good lenses, good lighting, and good fixturing. Each time you take one of these away, you reduce your chance of success by 24.25%.