determining aerial camera XYZ for convenient 3D model viewing in oblique photography

[angus absolute]

so here's the deal:

 

- working on a large energy plant construction project with some nice 3D models available

- once a month we receive updates of aerial, non-georeferenced photography (from a helicopter, not overhead, mainly oblique)

- the client has requested that the 3D models of the plant be dropped into some of the photos.

 

firstly, is there an "align" type tool that can be used to match points on an oblique-angled photo with a model thus determining a camera location and viewing direction?

 

if not, how can we do this manually?

 

my thoughts so far:

- known distances can be measured in the model and compared to distances measured in the photography as it appears

- photo distances can be scaled so that dimensions match in one direction, leaving us with a "ratio of discrepancy" (or vectors?) for each photo. can these vectors be used to determine camera xyz and viewing angle?

- the photo would need to be dropped in to a ucs determined by the results of those calcs (?)

- it would all only make sense from a single camera xyz / viewing direction. couldn't be rotated and maintain perspective.

 

ground levels can be assumed level for this exercise.

 

 

 

anyone tried to do this before?

 

my brain is frazzled and I feel bad spending too much company time on this when I'm not sure I can achieve a result!

 

thanks,

angus.

[Organic]

Personally I would just be scaling the photo (via fences or other visual features) to the known surveyed distance.

 

If that is not good enough then you would need to look into ortho rectification which I've never done although I think is fairly technical.

[angus absolute]

yep. but i want the model to fit the photo, not the other way around. i don't think that would be classified as ortho-rectification. coordinate systems are really irrelevant for what I want to do, it's just a visual.

[eldon]

A single photo has perspective, whilst the model does not. There is a problem immediately when trying to match. You have to distort one or the other.

 

You should be able to see from where the photo was taken by comparing lines of sight.

[Tyke]

If you are using MAP or Civil 3D you could try using the ADERSHEET command on a copy of your file. You would need to set up some control points from your photo that you would then transform your model onto.

[BIGAL]

A crude way try Vpoint 1,1,1 as a start then a predefined zoom factor play with the xyz values in vpoint.