Overlapping Objects

[diarmaede]

All,

I am working on code which differentiates bright objects in a camera's field of view.  Programming in LabVIEW, I have so far been able to detect and enumerate separate objects using LabVIEW's built in function.  My next challenge will be to differentiate between overlapping bright objects., objects which will cause contiguous bright pixels with no dark areas in between the objects.

I can identify static points which will always show one specific object, and I can always take my light intensity measurement points at these static points.  Is there a way that I can programatically have LabVIEW divide a group of contiguous bright pixels into two or more separate objects?  Or will LabVIEW's built-in functions always identify a group of contiguous bright pixels as one object?

If I cannot program LabVIEW to split contiguous bright pixels into multiple objects, I can figure out another way to code the functionality I want.  But if there is a way to have LabVIEW return an object array that already has differentiated the objects as I want, it would make my life a bit easier.

Thanks in advance.

[MoviJOHN]

Skeleton

[diarmaede]

Thanks for the response.  Do you know where I could find an example showing how the use the Skeleton VI?

[jcabrer]

It is called Watershed in LabVIEW NI Vision.

[diarmaede]

I was playing around with Skeleton, and I don't think it will work for me, but I could be wrong.

I am attaching an image that approximates what my cameras will be seeing.  What it shows is four equally-sized mirrors tilting up from the ground. 

Any thoughts as to how I could get LabVIEW to differentiate these into four different objects?

Thanks.

[BruceAmmons]

You might be able to do it with IMAQ Separation, but the number of erosions would be quite large.

If the mirrors are going to stay in the same place, I would consider manually defining the separation and saving it.  You could use a paint program to create labels (each area is a single intensity) from the original image.

Bruce

[falkpl]

Have you thought about pattern matching?  edge finding and connecting the projected edges?  You can look at the contour matching examples under the machine vision palette.