32'' Millennium Falcon photogrammetry.

[devi1sdoz3n]

A long time ago I've posted a thread about reconstructing the 32'' Millennium Falcon studio model.
I've finally caught enough free time (and unearthed enough reference photos) to work some more on it, and here are some results.
The preliminary reconstruction of the top is done, I'm working on the bottom (a bit harder, as there are less photos available).
This project will take a long time though, as I have little free time. I will post more as I go along.

 

[LrdSatyr8]

My question is... when they did the scan, they used a white light scan... why didn't they just do a 360 Laser scan of the entire ship in .0001 increments? I can understand using the hite light scan for the paint job, but the minut details would have been picked up alot better if it was laser scanned would it not?

 

[dsp5500]

SLS scanning trumps laser scanning in both speed and precision.

 

[imurme]

Very nice devi1sdoz3n, what are you using to do the photogrammetry? I am interested in doing this as well.

...would you mind a favor? I know your scan is a work in progress but could you post high-rez shots of the scan directly from the top and bottom in orthographic views (no perspective)? It would would be ever so helpful in working up my plans...

I look forward to seeing you complete this!

imurme

 

[LrdSatyr8]

I'm sorry, I'm just not convinced after seeing those "scans" above. They look awefully grainy and missing alot of data in alot of areas. Some areas seem to be just washed out. Maybe it's just me... I dunno!

 

[Lee S]

Phorogrammetry isn't a scan, it's taking multiple photos and working out a 3D shape based on them[/h]It's the same way you see depth, or that xbox add-on 'see' you movements.

L

 

[thorst]

I'd also be interested in which programs you use for this. Great first results!

Thorsten

 

[devi1sdoz3n]

LrdSatyr8, these are not scans, they are photogrammetry solutions from the highest quality photos I could find on the web. As close to the scan as I can get without having access to the real model, which is not at all likely...
This is not meant to be a detailed model like say Steve Starkiller makes, but as precise a reference as I can humanly make with the photos I've got. I can then use it to make precise plans, similar to what the other guys did with the 5 foot Falcon in the other thread.
The point is that this is not an interpretation, but a pretty accurate mathematical reconstruction. The grey photos show the actual extracted 3D shape, the ones in color have the extracted texture applied.
I am already seing interesting things, like asymmetries in the model where you wouldn't expect them.

 

[devi1sdoz3n]

Thorsten, I use Agisoft PhotoScan and am testing Acute3D Smart3DCapture free version. For now, Photoscan seems to be better for the general camera alignment, but Smart3DCapture blows it away for model reconstruction. Unfortunately, I can only look at the model with the free version. I'll probably buy the basic version of the software if I get results I am happy with (and for now it seems I will).

 

[vfxsup64]

Nice results!

I'm also interested in looking at other photogrammetry solutions (in addition to Photoscan).


But, the tricky thing about any image-based tool/technique is that the results can easily be misleading. Issues like lens distortion, exposure, grain/noise, depth of field and the degree of parallax triangulation can all multiply to either help a solve - or not.


I found with the ANH MF survey project that I needed 3 different data inputs, so to speak, to help corroborate any single tool's results:

- Donor kit parts (the most important, IMHO)
- 3D discreet feature point tracking (this allowed for wireframe lineups to confirm edges, corners, discrete features etc.)
- 3D photogrammetry (this allowed for surface comparisons and trends - i.e. the hull dome)

The three together, dovetailed nicely - at least for me!


But, of course, there are many different ways up a mountain...



Regards,

Andre

 

[LrdSatyr8]

Ok... now I understand a bit better... basically you're stitching together all the different hi-res photos of the Falcon to generate a 3D model right? You know I've got a bunch of links to a huge collection of actual hi-res photos of the falcon... from photos taken of the actual 32" that were taken while it was on tour in the Star Wars exhibits. If you're interested that is.

 

[thorst]

Thanks for the answer!

In principle it should be possible to measure a camera's lens distortion by photographing for example a raster grid like on a student's notebook and analyse the lines in the picture. Can such a calibration be done with existing software? I assume that if the lens parameters could be fixed for the reconstruction it would lead to much more stable results...

Thorsten

 

[dsp5500]

Most lens distortion can be corrected using the Brown–Conrady's distortion model. Stay away from wide and fish eye lens. You can also calibrate as you describe and save calibration files per lens.

 

[imurme]

Ok... now I understand a bit better... basically you're stitching together all the different hi-res photos of the Falcon to generate a 3D model right? You know I've got a bunch of links to a huge collection of actual hi-res photos of the falcon... from photos taken of the actual 32" that were taken while it was on tour in the Star Wars exhibits. If you're interested that is.

I would be interested in these photos as well, could I get copies? I'll PM you my email.

imurme

 

[devi1sdoz3n]

LrdSatyr8,
the more hi-res photos, the better the reconstruction.
So I'm definitely interested.

 

[devi1sdoz3n]

Thorst,
yes, it is posible to calibrate the lens, but I am using images I collected from all over the net so I can't do that. Right now I am using the largest images I have (around 3000 pixels on the longer side is the lowest I am willing to go for the pictures of the whole Falcon) with camera settings in the Exif. So I am depending on the software to deal with the lens distortion. Seems to be working pretty well so far.

 

[devi1sdoz3n]

Andre,
of course, you are right. I will have to check and double check everything in more ways. This is just the first step.
It is actually your wonderful work with the 5 foot Falcon that inspired me to get back to this (I like the 32" model proportions more - heresy, I know - that's why I'm working on it- And it's a challenge to see if I can do this to my satisfaction)

Chris

 

[devi1sdoz3n]

Here are some tests of the bottom side.

ChrisView attachment 440901View attachment 440902View attachment 440903View attachment 440904View attachment 440905

 

[imurme]

Here are some tests of the bottom side.

ChrisView attachment 440901View attachment 440902View attachment 440903View attachment 440904View attachment 440905

...seems your attachments did not work? could you try posting them again...

 

[devi1sdoz3n]

Hm, don't know what happened there, they were visible for a time.
Here goes again.