www.ssontech.com

matix · Joined: 08 Nov 2006 Posts: 32 Location: Savannah, GA USA

Just invested in Syntheyes2007. What a great product!

I'm doing some tests on footage, and i've come across a shot that I cant really get to work. There is no floor seen during the shot so the tracker I guess is relying on the walls. The auto-solve gets it ballpark, but it floats a little and I cant find good points to establish scene geometry.

As a test, i'm trying to place a simple box on the table, to see if I can get it to move with the footage like its sitting in the scene. See JPG below.

The shot pans and turns. If anyone is interested, i'll post the jpgs i'm using for the shot. I appreciate any ideas, etc!

gordonrobb · Joined: 17 Mar 2005 Posts: 281

I would love a go at it if you have the jpegs.

matix · Joined: 08 Nov 2006 Posts: 32 Location: Savannah, GA USA

matix · Joined: 08 Nov 2006 Posts: 32 Location: Savannah, GA USA

gordonrobb · Joined: 17 Mar 2005 Posts: 281

Had a look, the problem is that there are virtually no trackable features.

Everything is either a reflection or a point between two moving objects.

Wouldn't even know where to start.

matix · Joined: 08 Nov 2006 Posts: 32 Location: Savannah, GA USA

Yeah, I was running into the same problem. It also doesnt help that its DV footage.

I appreciate your attempt at helping me solve it. Anyone else have an idea at how to approach this?

LFGabel · Posted: Wed Nov 15, 2006 7:31 am Post subject:

There's tons of trackable features here. If the autotracker get it into ballpark territory, then the shot can definately be solved. I'd do this one manually, and/or heavily remove/refine the auto trackers. Where the blinds cross the window dividers is a good place to start. Once you get a solve that works, lay a tracker on the table near the end of the shot where you can see light and shadow, and track that for 8-12 frames (or as long as you can accurately do it.

If you are not masking out the people and anything else that moves, then there's really no way you'll get a good automatic solve. Do it manually. In the end you'll save time and get a better result.

matix · Joined: 08 Nov 2006 Posts: 32 Location: Savannah, GA USA

ssontech · Posted: Wed Nov 15, 2006 11:10 am Post subject:

This may mean that you still haven't got a good solve due to the problems discussed, or that you haven't set up a good coordinate system if you have. If you want to put something on the table, obviously the best coordinate system is going to be one aligned with the table itself---which may be difficult if there's nothing particularly trackable. You're almost certainly going to have to put at least one supervised tracker onto the table itself to use as a reference.

gordonrobb · Joined: 17 Mar 2005 Posts: 281

Doh! Don't know why I didn't think of that. Guess I was asleep.

matix · Joined: 08 Nov 2006 Posts: 32 Location: Savannah, GA USA

LFGabel · Posted: Fri Nov 17, 2006 5:39 am Post subject:

The trackers on the table don't need to be there all the time. If you have other accurate trackers elsewhere that take over when the table trackers get obscured, then it should be fine.

ssontech · Posted: Mon Nov 20, 2006 12:04 pm Post subject:

Just to clarify. Lee is absolutely correct---the trackers on the table don't have to be there all the time, just long enough to get a good solve on their position. The "taking over" part---you need to have an adequate collection of trackers throughout the duration of the shot, they can come and go. They should be spread around the images, not just on the table. You only need the 3, part of the time, and not even necessarily at the same time, to set up a coordinate system. You do NOT need a separate set of trackers for the coordinate system, when the main 3 are not visible.