If it was an image, not really, Henk.
Think of the screen as an X and Y coordinate system. In my imaginative coding thoughts it would sound like this: if the mouse is positioned and hold down at x=450 and y=213 (as screen pixel values) and if moved and increased value x, do this, if y is decreased to something else, etc.
If it was a little dummy model, things are differently. I preferred to answer the initial question with the image idea, as I remembered you other question about “how to set an logo in the GUI in the right lower corner. Edit: Rick confirms that it was a polygon Proxy, with that it was somewhere in 3D space, not stationary as in “Image on Foreground”./edit
To do that with a foreground object (or for better readability of the scene as back-ground object) seems way less problematic than to have a model connected to the camera. Which would cause trouble while changing the filed of view, the lens distortion (physical render) or by just switching from one camera to the next).
Solutions should hold less new problems than they solve, right? based on that, my reply above.
BTW, I remember that some people used a while ago software that places a “layer” over the current running app, so one is able to paint/draw on top of the GUI. How much this works these days against the GPU power, I have no idea. I try to keep the demand on my machine as low as possible. But that might be different for everyone, as I think that we have very fast machines these days, but some even complain while having 100 times the horse power than ILM had doing the first three Star Wars movies.
Again, I would do it in post. ;o)
All the best