How do I process realtime images to generate a control signal

[nellerin]

Hi, hopefully someone can help get me started on this problem...

My ultimate goal is to be able to:

1.  Capture an image

2.  Process it to locate the position and orientation of an object within the image (I am using 2 different size circles drawn on the object which should be easy to locate.  The center of the large circle is the center of the object.  Drawing a line through the centers of the two objects gives the orientation)

3.  Determine the error from this data to a required final position and orientation

4.  Output a callibrated control signal to manipute the object one step closer to the desired position and orientation

5.  Repeat steps 1-4 until I get to the final position and orientation

Sounds simple.

But I am new to Labview and NI Vision assistant, and haven't yet been able to integrate the two.  Does anyone have an example Labview vi that would create a loop that captures an image, processes it in some fashion to extract some positional data, outputs a control signal and repeats the loop until the user stops it.  If I can use this as a model, hopefully I can figure out the finer details.

(Note: for test purposes I am using a USB camera, with drivers that are allowing me to capture images in Vision Assistant for processing.)

Any information that can be thown my way would be appreciated!

Neil

[muks]

Use the initialize command and cloase command outside the while loop.For the acquire u need to define a buffer do that but place the imaq create out side the loop.Now if u have connected the camera properly u should be able to continously acquire the image.If u need further details let me know the camera u r using.
Now for the next part of your programming u can do 2 simple things.One is to go with pattern matching where u need to define the template and most importantly the the output will give u a cluster specifying the orientation.U can use it for further processign or u can use a co-ordinate system which is similar to pattern matching.Start acquiring images first the rest is pretty easy.

[nellerin]

Muks,

Thanks for the quick reply.  Your advice was helpful.  I was able to enumerate the two cameras that I have, and can select between them with the image aquired within the loop.  Works well.

Over the coming days, I will work on the image processing to see if I can extract the positional information.

Neil