Matching analysis in case of LCD

[anusree]

Thanks Analog,

I will try with the help of available Examples.

With regards,

Anusree

[RManion]

Hi Anusree,

The use of a coordinate system is pretty straightforward, but if you were looking for an example that uses this process, I do believe the Fuse Inspection Example that ships with Vision uses a coordinate system.  Additionally, there are most likely several other examples within this folder that might be helpful!

It can be found at:

C:\Program Files\National Instruments\Labview X.x\examples\Vision or via the Example Finder

The example I speak of is in the applications directory.

Good luck with your application,

Robert

[anusree]

Hello,

The problem has not fixed.

In the vi the image is connected to the imaq find coordsys (rect) and the image out from this is posted . then the image is connected to "extract image" , setting the optional rectangle as the LCD dimension.

When we run the vi if the device is deviated the complete LCD image is not extracted. how can we extract the image .

With regards

anusree

[anusree]

The deviated image is attached

With regards,

anusree

[DJ_L.]

Hey anusree,

I believe an answer to your last question has been answered on another one of your other discussion forum posts <a href="http://forums.ni.com/ni/board/message?board.id=200&message.id=7132">Extract image using coordinate system</a>. If you have any other questions related to this original post, please post them on this discussion forum thread, otherwise please do make a new thread about a new issue. Thanks for your cooperation in this matter and let us know if you have any furhter questions. Thanks and have a great day.

Regards,

DJ L.

[chilly_charly]

Hello Anusree,

I had not much time to comment the vi I proposed before leaving.

I'm a bit surprized you had problems with positionning : that was exactly what the vi was addressing :

before doing the subtraction, the screen image was located. A further step could be added to scale the image if there is also some distance variation (resulting in a difference in apparent size). Even a 5 mm move should have been compensated.

If you still have problems, you could post some example images.

[anusree]

Hello CC,

I was trying to solve the above LCD matching using pattern matching but sometimes it works and sometimes  result fail for the same passed LCD because unable to match scores . The particle matching by subtracting image works fine(earlier sent vi) but if there is slight shift in the image than the tamplate it gives result fail. Because extract rectangle for both template and the captured image is same.As you suggested to do scaling how can I do scaling to locate the exact LCD rectangle in the captured image from the camera and the template.

regards

Anusree

[chilly_charly]

What I did previously was to threshold the image to get the whole LCD display silhouette and locate the top and left most extreme values. Then use these values to cut a rectangle of size equal  to the LCD display, assuming that the display size (and orientation) was always the same.

Now, since your images are so different, you will have some additionnal work to do with particle analysis :

1/ locate the LCD display
2/ Find its orientation (max Ferret diameter orientation)
3/ Rotate the image to correct the orientation.
4/ Locate again the LCD on the rotated image (now correctly orientated) and find the bounding rectangle
5/ Extract the corresponding rect on the rotated image and re-sample the image to have standard pixel number in width and height
6/ Process as before by image subtraction.

But I think that a much better solution would be to improve your setup in order to have the LCD screen always perfectly positionned under the camera. Remember that before image analysis, improving the image acquisition is always a good idea.