Log in or register to post comments

Frame markers as multi image targets

May 31, 2011 - 9:11pm #1

I want to be able to look-and-track all around in my room. In order to do this I have to make a multi image target system and place trackables across all walls. But the viewing distance will be relatively big so an A4 format won't do. And sticking huge "stone" or "woodchip" pictures (yes, I can make my own, I know) to the wall doesn't look good. It would be better if I can take normal paintings and create a frame marker around it. I can split the frame marker up and print several pages, each containing it's own part. That way I can make the trackable much bigger with good resolution. And the picture itself will look good because I can use any "painting", whether it it makes a good trackable in itself or not.

So I want to be able to use frame markers in a multi image target system, instead of image targets. Can you put this in a future update?

Re: Frame markers as multi image targets

June 29, 2011 - 7:06pm #5

I found another reason why not to use image targets in a large distance multi-tracker environment.

If an image target brought into view it takes much longer to detect then when a frame marker is brought into view. In an environment where the markers are coming into and out of view at a rapid rate, it is important that marker detection (and tracking once detected) is very fast.

Image target detection is just too slow on current hardware but frame markers can do the job just fine. Image targets are superior for close up tracking but not suitable for large distance scenarios where the markers come in and out of view at a rapid rate.

Re: Frame markers as multi image targets

June 2, 2011 - 7:44am #4
Quote:

But can't you at least make the multi frame marker system to work so I can experiment with it?

Sorry, it's not quite that simple :) I'll pass on the request though.

It is something you can do yourself however, by relating the pose matrices for each frame marker. All you need to do is pull the markers into a unified coordinate system. Again, this thread should help:

http://ar.qualcomm.at/node/2000410

- Kim

Re: Frame markers as multi image targets

June 1, 2011 - 8:54pm #3

Yes, the entire frame marker must be in view. So I want to place more markers in the room than I would normally do when using image targets. This might increase computational power required but since the frame markers itself require less computational power then image targets, it might cancel each other out.

The aim is that there are always a few frame markers in full view, so partially blocked frame markers should not pose a problem.

The problem is that I cannot get a clean 2d representation of a normal big painting. I cannot scan it. If I take a photo, it will introduce perspective errors and perhaps some lens distortions. Maybe I can correct the perspective/lens errors in software before I make an image target out of it, I'll have to look at that.

But can't you at least make the multi frame marker system to work so I can experiment with it?

Re: Frame markers as multi image targets

June 1, 2011 - 6:57am #2

It wouldn't work very well with the current frame marker implementation. Unlike image targets, the entire frame marker must be in view in order to track. Multi targets work well because you can smoothly transition from viewing one image target to the next, you don't need both targets completely in view.

If you had a set of paintings around your room that contained strong features for tracking this might work. You would want a clean 2D representation of each painting to feed to the TMS system.

- Kim

Log in or register to post comments