OCR: urgent

[Sena_Scott]

Hi,

    Scenerio: My robot have to use the images it acquire of the surroundings and LabView will have to be used to perform OCR operation on these images and upon spotting the target(for example: the target is No:1)  wanted, it wil then cause my robot arm to push the "number 1"..  So i waswondering

      how does i actually use OCR? what i did in NI vision currently.. Firstly, i Acquire the image then i threshold it, regarding thresholding( i was wondering how do i know what is the min and max values i need to set for the RED, GREEB and BLUE??  Then After that i WEnt on to OCR..  I kind of unsure what values i need to put in most of the settings.. I Knew i need to set the character file which i draw a ROI around "number 1" and saved it..  Then i use the NI Vision Assistant >> Tool >> Create LabView VI...  Upon running the program , i'm not sure how do i know the program is functioning according to the way i wanted?

Thanks, this is very urgent...

[Adam_H]

Hi Sena_Scott,

Starting with your question about how to set the min and max values for a threshold.  This is kind of a trial and error.  What values you select is going to depend on your image.  My best recommendation is to try different values and see what seems to work best.

As for how to use the OCR I am going to assume you don’t have a charter set file.  You will first select the “New Character Set File.”  In the following menu you will highlight the region of interest (ROI) you would like to recognize.  The program will then recognize the chanters but you must tell it what they are.  On the right side of the screen there is a box called Training.  Select “Train All Characters” from the list.  You can then type what the charters are in to the “Correct String” dialog box.  Click train and then you can save the file.  This should set your OCR up correctly. 

When you run the program does it seem do what you want it to or not?  Does it give you an error or simply not work.  Your best bet is to test it and see how it responds.

Thanks,

Adam H