Jump to content

VIZ Camera Matching-Photo-simulation


lrose
 Share

Recommended Posts

I think this is a question for architecture-types.

My partner and I have been working on creating legally defensible photorealistic-simulation for public meetings. The problem revolves around the import of spacially accurate, real world coordinate AutoCAD/Inventor 3D models into VIZ. We have a high level of confidence surrounding our real world coordinates. Cameras and camera target locations are not faithfully reproduced in VIZ, and always need to be tweaked by the artist's hand and eye.

This is unacceptable when dealing with legal documents.

We have tried utilizing PhotoModeler in the process, but still find issues when trying to import the camera information into VIZ's idealized virtual camera representation.

 

Has anyone had success with this process, or is there a package out there that better coordinates with VIZ? Or maybe there is another package besides VIZ that will accomplish what we want?

Link to comment
Share on other sites

not sure if these are the specific methods used by HD but just googled verified camera matching and found the following, might be worth a browse:

 

http://www.cc.gatech.edu/~dellaert/workshop/final/zhang.pdf

 

http://www.robots.ox.ac.uk/~vgg/publications/papers/zisserman99.pdf

 

http://www1.cs.columbia.edu/~allen/PAPERS/cviu.ioannis.pdf

 

i'm sure there is probably many more papers too,

 

James

Link to comment
Share on other sites

  • 2 weeks later...

well, i have done a lot of this...

 

but, photos are not perspective correct.... they will will have to be adjusted to the model..... pans are the worst... they are helpfull for sight lines but not for realism..

 

not sure what you mean as "unacceptable" tweaking is absolute must no mater what the legal issue is...

Link to comment
Share on other sites

well, i have done a lot of this...

 

but, photos are not perspective correct.... they will will have to be adjusted to the model..... pans are the worst... they are helpfull for sight lines but not for realism..

 

not sure what you mean as "unacceptable" tweaking is absolute must no mater what the legal issue is...

 

We have made some good progress on our technique. The parameters that seem to have made the biggest improvements are insuring that our reference object in the first photo symmetrically fills the frame as much as possible. This alows our control points to be spread across the maximum lens area. Then reducing the RMS error as much as possible for each point.

 

The other issue seems to be having very well defined distance measurements from which to work. These include measurements from the camera to the "field target(s)".

 

Our next step is to tighten-up our field techniques for the large-scale conditions that we usually need to operate under. Accurate distance measurements are the biggest issue in this case.

 

Thanks, all, for your interest and your help.

Link to comment
Share on other sites

interesting...

 

do you take photo's with references (like story poles) at the site.

for example to show exactly where the building corners are.

 

i used to work in a area where a vizual simulation was part of the permit process... it could make or break a project...

 

i should know but don't what is rms error??

Link to comment
Share on other sites

Sorry I wasn't more specific.

 

For our permitting photos, many times there are no existing structures in the area, so we have devised our own large target(s) to place in the field of view---sometimes at various locations, multiple shot, from a stationery camers location. Then we blend the shots to give the impression of mutiple targets in one shot. If there are existing structures, we try to determine accurate dimentions for them for reference.

 

RMS (root mean square) error is calculated by our application, basically giving a determination of how accurately the control point locations match where the software believes they should be located in the photo. This position can be modified after initial placement in the scene, with a new RMS error value being recalculated to show improvement, or not. The smaller RMS error, the better.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share

×
×
  • Create New...