Strange behaviour of the IMAQ Read LCD function

[Garcas]

Hi,

I am developing an application in LabVIEW where I need to obtain the reading of an instrument that does not have any programmable interface. The solution I am evaluating now is to use an image acquisition board and the IMAQ functions to recognise the instrument reading. The instrument�s display is an LCD.

The problem is that sometimes the IMAQ Read LCD function fails to recognise the instrument reading. It is really strange because it recognises the illuminated segments properly but then interprets them wrong, as it can be seen in the attached file (Failure Example of the IMAQ Read LCD function JPEG Image).

The file shows an image acquired with the camera showing a 99.8 reading in the instrument displa
y. The Segment status output of the IMAQ Read LCD function is correct, but the output value (�Salida texto� indicator) is wrong. It is confusing a 9 with a 4. Strange, isn�t it ?

I am not an expert in Machine Vision technologies, but I thought the most difficult part of the process would be to establish which segments in each digit where illuminated. Once this is known, I thought it would be easy to recognise the number.

I would like to know if there is something that I am missing (I guess it its), or if it is a bug of this function, or... anymore ideas about this?

Thanks a lot.

PS: I�m also enclosing the files of the program that I�ve used to evaluate this problem (Program files.zip).

[BruceAmmons]

It looks like it is reading the segments of the LCD correctly, like you said. The problem is the interpretation of the pattern.

My best guess is that the routine assumes that the bottom horizontal segment will be on for a nine. When it doesn't find it, it does a closest comparison and comes up with a four. My solution would be to convert the cluster of LCD segments that comes out into an array of booleans, and convert that to a number. Wire the number to a case statement, and do your own interpretation of the LCD segments. It shouldn't take long to figure out which segment pattern corresponds to each number.

An alternative option is to use OCR, but that seems a bit extreme for this application.

Bruce

[Rlima]

  Hi  Bruce...

  How are you  today...?

  I'm working late today, I have to finish a code for friday morning, is my time line, 

  I want to do a vision inspection in a display,  I just found this code, i downloaded did not work

  some vi are missing, "vi is not executable"  I',m pretty sure is a software version I'm running

  LabVIEW 7.1  NI Vision  7.1 and VAS 8.5.    I know this is for 7 segments display, I will need to

  do something in another project that I'm quoting..

  Attatch you will find and image of the display.  I ran  OCR,  seem to work OK for me but I have

  a little trouble with the letter "M".. OCR separtes in 2 or 3 characters,  I think I can manupulate the

  string   3 charactes  = to "M".

  Thanks for you help...

[BruceAmmons]

Doing pretty well, thank you.

It looks like OCR should work fine on this display.  Here are a few observations:

The space between the two parts of the M are about the same as the space between the F and the M.  If the M space was smaller, I would recommend adjusting the spacing parameter that adjusts max space between components of the same character.  That still might help you out to join the character pieces.

Each section uses a different font, so it might be useful to define a separate font for each section, if your analysis can read each section separately.  If you have to read the entire line at once, you will just need to merge all the fonts.  There is more possibility of error when you mix fonts.

Bruce

[Rlima]

        Hi Bruce..

        Thanks for you quick response.

        I will consider your comments on my code..

        I'll keep in touch..   Take care...

        Best Regards...