Testing LCDs for defects

[David Cabernel]

I've successfully tested a 7 segment LCD display for all the correct segments coming on for the correct message. My next problem is to test for:

1) Marks on LCD, either dirt on the LCD or defects in the LCD.
2) ghosting, segments not supposed to be on, slightly on.
3) Contrast problems, ie. some segments not as dark as others.

My biggest headache with this display is that it is backlight poorly. The sides of the LCD are much brighter than the middle. Any thresholding I do does not work in all parts of the screen. I could break it up into
different regions, but the dividing line between light and dark is very gradual.

Any tips and suggesions that have worked for you would be appreciated.

[PdB]

First of all, I have no experience in vision applications, just have some experience in measurements-applications.
Can't you eliminate the background effect by measuring the intensity at non-segment positions along the display?
Fit a 2D curve for this 2D-pixel_array (to mimic the non-ideal background lighting) and subtract it from your original picture. This should clean your signal, in theory.

[Mike H2]

You might want to try using a LUT to clean up your source.

[Labviewguru]

Lighting is very key to any vision application. I had to deal with a situation where I was imaging an LCD under a cover. We were yet unable to solve the lighting, but did find a solution that we are going to try out. The lighting solution is an Diffuse Axial Illuminator. They are pricey, but not when you consider the software hours required to make up for a poor image. This illuminator may be what you need. Edmund Scientific makes a great one.

Good luck

[David Cabernel]

Thanks for the suggestions. I got looking for marks in the LCD or dirt behind working by scanning small areas at a time and using the Quantity VI to calculate the mean value of the pixels in that area. Then I set
the upper value of the Threshold function dynamically by setting it to 10 - 15% of the mean value. In this way, anything significantly darker than the area scanned gets counted as a mark in the display, and it
doesn't matter if it is in the bright or dark section of the LCD.

It seems to work well, a small pencil mark on the glass would get detected even if it was not that noticabe just looking at it. I divided the LCD up into about 120 sections and it takes about 8 seconds to scan the
entire LCD. I am currently scanning a blank LCD, but
I think I will do more scans when part of the display has segments on as some LCD defects only show up when other parts of the screen are on.

- dave