Go Back   Foundation 3D Forums > Creative Works > Works in Progress
Register FAQ Members List Calendar Mark Forums Read

Works in Progress Critique works in progress, 2D 3D and animations

Reply
 
Thread Tools Display Modes
Old 5th Nov 2009, 09:35 PM   #1
McC
Webmonkey
 
McC's Avatar
 Gallery
Blender Photoshop After Effects
 
Join Date: Dec 2007
Location: MA, USA
Posts: 609
Default Star Wars - His Imperial Majesty's...

I haven't posted a new project to the boards in a long time. The last several were abortive, to boot. Here's hoping this time will be different.

It's been a long-standing aspiration of mine to do up an Imperator II Star Destroyer with close-to-screen accuracy. Obviously a lofty goal. I made one or two attempts at it back in my LightWave days, and I think I got as far as the engine bells before something else caught my interest and I moved on. The other roadblock was/is my over-reliance on orthographic reference images, for which there are none of the ISD (the diagrams in the Essential Guide to Vehicles and Vessels are deplorably inaccurate).

Thanks to some encouragement from my fiancee and from fractalsponge, I've decided to take a crack at it once again. I don't plan to be fast, and I may well stay in the low-poly layout stage for quite some time until I (and you!) feel that I've accurately captured the correct shape.

To that end, here's my first chunk of stuff, done across two nights (two nights? it took you that long to do this tiny amount of work?). My major focus has been trying to get an accurate slope angle for the dorsal and ventral main hull surfaces.

I started with the only known (to me, anyway) full-on, perspective-corrected ventral view of a Star Destroyer to get my basic arrow-head shape. This is the Devastator (Imperator I) hull, so it may not be 100% identical to the Avenger (Imperator II) hull, but for gross proportions, I think it should match.


The next step was to determine the slope from front-to-back. Initial measurements from photo references suggested a slope of about 3° for the dorsal surface and 1.5° for the ventral. I tried this out and it looked ridiculous; I was pretty clearly suffering from some perspective distortion.

It finally dawned on me that while I may not have good reference of the lateral view, I did have a good reference image of the back of the destroyer. Since I was pretty confident about my top-down silhouette, I could measure the slope from midline to edge to define the slope along the length. The green lines in this image denote verification that I had properly corrected the skew in this image.

With a bit of MaxScript-driven trig, my slopes now match the reference image.

The next trick was figuring out how far apart the two shells needed to be. I once again turned to the same back reference image. This time, I measured the proportional distance between the peaks of the drive bay along the centerline and compared it with the height of the trench. In this case, perspective distortion will alter my measurement somewhat. However, it shouldn't be extreme enough to throw off the measurement by such a degree as it will be noticeable.


The gap between the trench vertices is thus 45/589*(vertical separation of apex vertices). Here's the roughed-out shape:

The terraces (superstructure, towers, etc.) have a slightly different slope with respect to the hull's center-to-edge slope, which seems to be about 12.1°. There are three full tiers, some of which are subdivided. Fortunately, perspective distortion decreases as you recede from the camera, so the front-on reference image I have should actually prove fairly useful. We can compare the trough height where the terraces begin (approximately at the minor notch in the dagger) with the terrace height measurement to get their approximate dimensions.

Here are some large planes that will end up getting cut into the terrace pieces.

I need to figure out exactly where the terraces are placed front-to-back. However, that will have to wait until my next opportunity to model.

Commentary, criticism, and advice very welcome!
McC is offline   Reply With Quote
Old 5th Nov 2009, 10:51 PM   #2
Marty Miller
Staff
 
Marty Miller's Avatar
 Gallery
3DS Max Maya Vue Illustrator Photoshop Boujou After Effects Premiere
 
Join Date: Jan 2008
Location: Sweden
Posts: 1,311
Default

Mate looks your off to a good start. |Will be looking forward to seeing this project progress.
Marty Miller is offline   Reply With Quote
Old 7th Nov 2009, 11:32 AM   #3
McC
Webmonkey
 
McC's Avatar
 Gallery
Blender Photoshop After Effects
 
Join Date: Dec 2007
Location: MA, USA
Posts: 609
Default

Thanks, Marty.

The next phase in figuring out the shape is in the front-to-back length of each of the terraces. Though perspective distortion is going to a problem with every photo, those taken from the side should be sufficient when measuring coplanar elements, like the terraces. They won't be good for establishing the front-to-back slopes of the terraces, unfortunately, but we'll cross that bridge when we come to it.

There's already a problem evidenced by this image. The ratio of terrace heights is inconsistent with the measurements taken from the front. Since the measurements here are likely to be more accurate (due to all elements measured being coplanar), I'm going to give preference to them.
Code:
 Measurement      Old Ratio        New Ratio
----------------------------------------------
T1:T2::T2:T3    65:48=1.35417    106:62=1.7097
T2:T3::T3:T4    48:30=1.6         62:58=1.0690
At first blush, this is a pretty significant problem. However, I tried scaling down reference points place with the correct relative offsets and when I line up the top of the T1 reference point with the top of the T1 that I created previously, and also line up the top of T2 reference point with the top of T2, T4 almost lines up with the established top of T4 from before, meaning all I need to do is make minor adjustments to T3 and T4's vertical position to bring it all back in line.

Having done that, I need next to figure out where the terraces start. This is going to be tricky, since I'm going to be fighting perspective distortion the entire way. I have two side images that show the front of T1 relative to some feature on the lateral trench (either the small back notch, or the larger forward notch).

T1 is above the hull plane by 42.187m based on my current arrangement. The implication from the first image, then, is that the top of T1 should be positioned 64.599m forward of the inside front edge of the minor notch. The implication from the second image is that the top of T1 should be positioned 258.565m afterward of the inside back edge of the major notch.

Here are the now-corrected terrace planes.

Next step is to properly shape the terraces, especially accounting for their top-to-bottom slope.
McC is offline   Reply With Quote
Old 7th Nov 2009, 12:06 PM   #4
Tenement 01
Registered Member
 
Tenement 01's Avatar
 Gallery
LightWave Photoshop After Effects Premiere
 
Join Date: Dec 2007
Location: Glasgow, Scotland
Posts: 3,557
Default

Good luck with your SD. Looking forward to seeing updates.
Tenement 01 is offline   Reply With Quote
Old 7th Nov 2009, 03:47 PM   #5
McC
Webmonkey
 
McC's Avatar
 Gallery
Blender Photoshop After Effects
 
Join Date: Dec 2007
Location: MA, USA
Posts: 609
Default

Thanks, Tenement.

The slope along the dorsal surface (front-to-back) on my model is ˜4.45°. One of the side images I have measures a slope of about 4.8°, so it's pretty close to perspective-free in that regard. It also shows the terraces rather well. Guess what slope they have? About 4.8° So it looks like they actually follow the slope of the main hull. There's some deviation in this measurement (some terraces measure 4.6°, and others 5.2°), but the deviation is small enough that I'm willing to chalk it up to measurement error and perspective distortion.

The next trick is figuring out how wide each terrace is. I have a front view from which I can measure, but there's going to be perspective distortion. I also have a view of the destroyer from the front, looking sort of down onto it at an oblique angle. This is going to have even more perspective distortion. I decided to measure it anyway, too. I compared the terrace-to-terrace ratio of the two images and they actually lined up fairly closely.
Code:
   | Width 1 | Ratio | Width 2 | Ratio 
---------------------------------------
T1 |   269   |  n/a  |   568   |  n/a  
T2 |   331   | 1.230 |   683   | 1.202
T3 |   234   | 0.707 |   476   | 0.697
T4 |   247   | 1.056 |   511   | 1.074
The ratios between the ratios (hoo boy...) also gradually decrease as the terraces increase, which is precisely what you would expect with perspective distortion involved. So, I'm pretty confident about the ratios represented by the Width 2 numbers (from the front-on image). However, the trick now is to relate them to something that I already have modeled so that I can get a concrete value. That's where the vanishing point image comes into play. I can draw a line from the back corner of the forward notches to the vanishing point, measure the distance between the notches, and then scale that back through the whole image.

The yellow line is a dropped-down projection of T2's width. The height of the green line shows the height of the terrace in this image, and the angled yellow line is drawn from the vanishing point through the bottom of this line, and then out to cross the horizontal line that bisects the notches. This should tell us roughly where this terrace sits with respect to the notch. The width between the notches is 475px, and the width between the yellow line/purple line intersection and the midpoint is 217px. So, T2 should be 434px wide in this picture if it were coplanar with the back of the notches. What's important is the ratio 434:475 -- this, multiplied by the 3D width between the notches, will supply the actual width of T2, and from there every other terrace width. Without further ado:

Terrace 1 is tapers along its length, and also comes to an apex at the center line, so it's current overall shape is still incorrect. Both Terrace 3 and Terrace 4 extend most or all of the way to the back, and have a bit of a taper as they do so. Every terrace except Terrace 1 has its outer edge sloping at more or less the same angle as its inner edge (~4.5°).
Code:
 L1  L2 Ratio
--------------
107 372 0.2876
 29  93 0.3118
 88 340 0.2588
Averaging these out, one finds a ratio of 0.2861 between the short and long legs. Here's what that looks like:

A few notes about the terrace block-ins here:
  • Some (all?) of the terraces should slope outward as they descend toward the hull (they do not in this block-in)
  • Terrace 3 and 4 taper toward the back, as noted. In this block-in, I eyeballed it. I will revise with proper measurements.
  • I eyeballed the outer slope of Terrace 1; I plan to re-do this slope with a proper measurement.
  • Terrace 2 should come much further forward (relative to the hull notches and Terrace 1) than it does now, and so too should Terrace 4 (relative to Terrace 3). This is throwing off the shape. I'll have to investigate the reason for this.
That's all for now. Dinner time!
McC is offline   Reply With Quote
Old 7th Nov 2009, 08:44 PM   #6
McC
Webmonkey
 
McC's Avatar
 Gallery
Blender Photoshop After Effects
 
Join Date: Dec 2007
Location: MA, USA
Posts: 609
Default

Okay, minor crisis averted. I realized at the very end of the last post that something had gone horribly, horribly wrong with the terraces. The minor notch should line up with about the middle of Terrace 1. It, um, didn't. So, I scrabbled around to try and fix it, eventually working up this template to follow.
ref-

I retained all of my other measurements thus far (the 4.5° hull slope, the 12.1° terrace slope, the existing terrace widths, etc.) and got back to where I was at the end of the prior post, but with terraces that actually looked correct! I then proceeded to work on correcting the shape of the rear of Terrace 2 and 4, using this image (which, nicely, happens to be relatively well-centered on the section in question) to measure from the back of the ship and the outer edges of the terraces. The top-down angle of the terrace rear seems to align pretty closely with the angle the whole hull has, so I just retained that.

T3 and T4 both have extrusion angles of about 77°, while T2 and (probably) T1 have steeper angles of about 81°

T3 and T4 don't have any major overhang. T1 and T2 have hooded perimeters. T4 and T2 both hang out in the back, though. In the back, both T2 and T4 have a sort of swooping hood that sticks out into the neck, while the main bulk of the superstructure follows the 77/81° setup.

So, putting it all together:

Probably done for the night. Need to spend some time with the fiancee!
McC is offline   Reply With Quote
Old 8th Nov 2009, 09:26 AM   #7
doncha_magoso
Contributing Member
 
doncha_magoso's Avatar
 Gallery
3DS Max LightWave Modo Mudbox Illustrator Photoshop After Effects Premiere
 
Join Date: Nov 2007
Location: Morelia, Michoacán, México
Posts: 518
Default

woha looking promising

keep it up!! it's looking good
__________________
woha! veggie burrito! :D
doncha_magoso is offline   Reply With Quote
Old 8th Nov 2009, 12:29 PM   #8
DaveK
Registered Member
 
DaveK's Avatar
 Gallery
LightWave Poser Photoshop Premiere
 
Join Date: Dec 2007
Posts: 741
Default

I wonder if anyone building those models, imagined that their work would under go such scrutiny at some time in the future.
DaveK is offline   Reply With Quote
Old 10th Nov 2009, 09:44 PM   #9
McC
Webmonkey
 
McC's Avatar
 Gallery
Blender Photoshop After Effects
 
Join Date: Dec 2007
Location: MA, USA
Posts: 609
Default

Thanks, doncha_magoso!

Quote:
Originally Posted by DaveK View Post
I wonder if anyone building those models, imagined that their work would under go such scrutiny at some time in the future.
I doubt it.

The next step is establishing the neck/tower. The first thing I did was measure the front and back angles of the tower. Because of varying camera angle and perspective distortion, none of the images is entirely reliable. However, taken together, a pretty consistent aggregate result emerges.
Code:
           Front   Back    Dist.   Angle
-------------------------------------------
Image 1    13.5    53.5    Mild    Mild
Image 2    17.5    60.0    Mild    Moderate
Image 3    13.0    54.5    Mild    Mild
Image 4    11.3    55.0    Major   Mild
So, I feel pretty confident (at this point) in saying that the front slope is about 13° off of vertical (or 77° incline) and the back slope is a 54° incline.

However, the neck has a bit of a rise to it before it begins sloping in the back. I used the same images (where visible) to measure this rise in proportion to the height of T2.
Code:
           T2:NeckRise
---------------------------
Image 1       43:6   ≈ 7.17
Image 2       90:7   ≈12.86
Image 3      150:21  ≈ 7.14
I feel the least confident about the measurement in Image 2, so I'm going to estimate that the neck rises about 1/7.15th the height of T2 from the hull before beginning its slope.

I'll use the same ratio method to try to identify the height of the tower itself.
Code:
            T2:NeckHeight
----------------------------
Image 1       43:137   ≈3.19
Image 2       90:275   ≈3.06
Image 3      150:450   ≈3.00
Of the three of those measurements, I trust Image 3 the most. They are, of course, all fairly close.

Here's the composite reference image:

And here's what it looks like:

I haven't brought it forward at this stage yet, since I don't yet know how far forward it's supposed to come. I also don't know how wide it is, nor how far forward the back is meant to come.
Code:
         T2-T4|NeckL|NeckF|  L:T |  T:F  |  L:F
------------------------------------------------
Image 1    41 |  93 |   4 | 2.27 | 10.25 | 23.25
Image 2    69 | 214 |   7 | 3.10 |  9.86 | 30.57
Image 3   121 | 324 |  17 | 2.68 |  7.12 | 19.06

NeckL = Neck distance from rear apex to forward terrace intersection
NeckF = Distance away from aft peak to start of neck base
T2-T4 = Distance between outer corners of Terraces 2 and 4

Unfortunately, none of these are very consistent! I am most inclined to trust image 3, due to both its size and its relative lack of distortion, particularly in this area of interest. If I use the average of all of them, I come up with ratios of 2.68 (same as image 3!), 9.08, and 24.29. Here's what those look like:

One problem becomes apparent here -- the neck is sticking off the back. There are two possible reasons that spring to mind: the neck is not pushed far enough forward, or the angle of the aft section is too sharp.

I looked through the references in the hopes that perhaps the point where the neck intersects the hull angles forward slightly, but in the images where I think I see this, it's probably just wishful thinking. Pushing the back corners further toward the aft wouldn't hurt any of the existing ratios, fortunately. On the other side of the coin, the current hull shape does seem to match pretty closely with the reference images. The corners of the hull should hit right around where T4 begins sloping inward, which is exactly where it does on my model.

As a dirty compromise, I can just move the back of the whole neck structure forward slightly until it hits the hull correctly. It's not exact, but given the small distance involved, I'm not going to sweat it.

For my final trick of the evening, measuring the width of the neck. This is pretty straight-forward, thankfully. This back view shows T4 to be 287px wide from center to edge, and T2 to be 367px wide. When compared with their 3D measurements, they have meter-per-pixel ratios of 0.47 and 0.49. So, they're pretty consistent. A simple measurement across the base and apex of the neck yields 87 and 45 pixels, respectively. The top is going to suffer from more perspective distortion, but that can be adjusted later.

Here's the "completed" neck rough-in:

If anyone spots any major issues with my measurements, or if the shape looks off, please do let me know. I have no idea if my approach is yielding accurate results. It seems to be (i.e. it looks more or less like a star destroyer), but I'm sure there are people who have looked at the same references images many, many times more than I have.

And now it's time for bed.
__________________
Then tells you when or next, than lets you compare. It's easy!

Finished: Ambassador Class | T'Varo BoP
WIP: U.S.S. Coronado | UEDF Illustrious Defender | His Imperial Majesty's...

Ashes of Alour-Tan | Tutorials | Blog
McC is offline   Reply With Quote
Old 11th Nov 2009, 08:45 PM   #10
McC
Webmonkey
 
McC's Avatar
 Gallery
Blender Photoshop After Effects
 
Join Date: Dec 2007
Location: MA, USA
Posts: 609
Default

Next up, the command tower itself.

I've got both a front and back image that feature the command tower in relation to T4. Both suffer from some amount of perspective distortion and/or camera angle.
Code:
            T4   Tower  Ratio
------------------------------
Back View   577   621   1.0763
Front View  529   514   1.0292
Not a pleasant disparity. The back view gets preferential treatment by virtue of the fact that the elements are closer together (back of T4 and back of tower vs. front of T4 and front of tower in the front view). Practically speaking, it means the bridge tower is either 292 meters wide, or 279 meters wide. From eyeballing other, oblique-angle photos, I'm more inclined to believe the 292 (which correlates with the preference it would naturally receive), so 292 it is.

The next trick is figuring out how tall it is. There are three measurements here that should all line up: the slope of the bridge (the bridge is actually a tiny protuberance from this section, but it's simpler than saying 'command tower'), the height of the outer edges, and the height of the middle section.

Assuming the bridge has the same slope on top and bottom, left and right, averaging all four angle measurements should yield a good value. On one image, I get an averaged measurement of 8.425°. On another image, with far more camera angle drift, I get an averaged measurement of 8.025°. Very close. Since they're in the same ballpark, and one image has less perspective distortion than the other, I'm going to go with the 8.4° measurement.


The next trick is figuring out how tall the thing is. I've used my go-to face-on image to take some different measurements (distance from T4 to T2, distance from T2 to the hull, the height of the tower edge, and the height of the tower maximum). I've also taken the same measurements (except the tower maximum, which isn't visible) from my go-to side-on view.
Code:
                         px      m      px/m
---------------------------------------------
Img1:T4-T2 (A)           72    29.03    2.480
Img1:T2-Hull (B)        123    66.41    1.852
Img2:T4-T2 (A)           99    29.03    3.410
Img2:T2-Hull (B)        163    66.41    2.454

This yields the following interpolations:
Code:
                         px      m      px/m
---------------------------------------------
Img1:Edge Height A       64    25.80    2.480
Img1:Edge Height B       64    34.55    1.852
Img1:Center Height A    142    57.25    2.480
Img1:Center Height B    142    76.67    1.852
Img2:Edge Height A       89    26.10    3.410
Img2:Edge Height B       89    36.26    2.454
At first glance, there's some pretty wild disparity in numbers that should be pretty close. However, look at the four edge height measurements. The average of Image 1's A and B edge heights (30.175m) is pretty close to that of Image 2's A and B (31.18m). A mere meter different. So, we can at least be reasonably sure that the correct number is in this range.

The next trick is to compare the edge heights with the assumed slope (8.4°) and see where that puts us for maximum heights, and which measurement (if any) lines up with that.
Code:
Edge Height    25.8    26.1    30.7    34.6    36.3
Slope          ---------------8.425°---------------
Width          ----------------292m----------------
Maximum Height 69.0    69.3    73.9    77.8    79.5
Based on our "maximum" measured peak height, the last two values aren't right. The lower three, however, all fall within the 57-76m range. I'm inclined to let the law of averages win this one and go with 30.7 and 73.9.

At this point, it's time to do a little cleanup on the overall shape. Perspective distortion and lens angle are starting to compound into issues that cause the proportional shape to be inaccurate, despite the measurements. Internal reference (i.e. eyeballing what goes where, relative to other bits) is the remedy.

Here's how everything looks, with the revised proportions from eyeballing plus the bridge tower:

C&C (particularly about the shape) very welcome at this point.
__________________
Then tells you when or next, than lets you compare. It's easy!

Finished: Ambassador Class | T'Varo BoP
WIP: U.S.S. Coronado | UEDF Illustrious Defender | His Imperial Majesty's...

Ashes of Alour-Tan | Tutorials | Blog
McC is offline   Reply With Quote
Old 11th Nov 2009, 09:00 PM   #11
Rob
Premiere Member
 
Rob's Avatar
 Gallery
LightWave Mudbox Poser Terragen Illustrator Photoshop After Effects Premiere
 
Join Date: Feb 2008
Location: Australia
Posts: 1,036
Default

ok and i thought i was a stickler for getting messurements right

I can ask a friend and maybe get you some more photo's of this.. can't promise though this time of year starts getting busy work wise for a lot of ppl around here and where he works.. = busy as hell..

I do like the work going into this though so far
__________________
- Rob
NDA's suck.. they take away your ability to go 'look shinnies'.. on the other hand they mean your working on intresting projects.
Rob is offline   Reply With Quote
Old 12th Nov 2009, 07:30 PM   #12
McC
Webmonkey
 
McC's Avatar
 Gallery
Blender Photoshop After Effects
 
Join Date: Dec 2007
Location: MA, USA
Posts: 609
Default

Thanks, Rob! If you can get ahold of those references without too much trouble, that would be most excellent.

Starting to exit the measuring phase and enter the eyeballing phase. The thruster width, secondary thruster housing width and offset, width of the large side blocks, and inset of the thruster cowl are all measured. Everything else, including various slopes, thruster separation, and so forth was all eyeballed.



I noticed several small shape issues with the underside of the hull, so that's also been corrected now, too. It had been looking too steep to me, and sure enough, it was steeper than the upper section by a bit.

I'm really starting to hurt for those last few outer terraces, though, so I think I'll tackle them next. I'm also thinking about fattening the bridge tower somewhat. It looks a bit too vertically squashed from this angle. That might also allow me to pull it back some, so that it doesn't come quite as far forward on the neck, which seems a little extreme right now.
__________________
Then tells you when or next, than lets you compare. It's easy!

Finished: Ambassador Class | T'Varo BoP
WIP: U.S.S. Coronado | UEDF Illustrious Defender | His Imperial Majesty's...

Ashes of Alour-Tan | Tutorials | Blog
McC is offline   Reply With Quote
Old 12th Nov 2009, 08:52 PM   #13
Treybor
Asylum Inmate
 
Treybor's Avatar
 Gallery
LightWave Poser ZBrush Photoshop Premiere
 
Join Date: Nov 2007
Location: San Diego, California
Posts: 6,357
Default

Wow. keep this up and you should have an accurate 3d studio model.
__________________
Keep Flyin!
https://keepflying.com/
That Serenity Tutorial Guy.....
http://www.foundation3d.com/forums/s...ead.php?t=1281

SubPatch Modeling Tutorial (Nudity Warning)
http://www.foundation3d.com/forums/s...ead.php?t=3445
Treybor is offline   Reply With Quote
Old 12th Nov 2009, 09:31 PM   #14
3D_CG
Contributing Member
 
3D_CG's Avatar
 Gallery
3DS Max Blender LightWave Terragen ZBrush Gimp Paint Shop Pro Photoshop Digital Fusion Premiere
 
Join Date: Jan 2008
Location: Spokane Washington, North Side in a quiet neighbor hood, Tell I start My Harley, Then its NOISY...
Posts: 4,941
Default

wow wow wow
__________________

"Remember The Vets who fought for our Freedom" ...
Intel I7 980 OC'D to 5.1 GHZ, Water Cooled, Gigabyte MB
132 gig ram, 5 HD's, TITAN Z Black , GTX 1080,
LightWave 11.6.3, 32 and 64 bit

http://boincstats.com/signature/-1/user/355618/sig.png
3D_CG is offline   Reply With Quote
Old 13th Nov 2009, 06:46 AM   #15
ChrisGFX
Registered Member
 
ChrisGFX's Avatar
 Gallery
Cinema 4D Photoshop After Effects Premiere
 
Join Date: Aug 2008
Location: Pulheim / Germany
Posts: 272
Default

awesome!
__________________
If you can’t explain it to a six year old, you don’t understand it yourself.
ChrisGFX is offline   Reply With Quote
Old 13th Nov 2009, 10:48 AM   #16
Sastrei
Registered Member
 Gallery
Blender Photoshop
 
Join Date: Feb 2008
Posts: 245
Default

Question related to the references thread over in "3D" - during modeling this, do you tend to find the ship comes out being narrower or taller than you think it should be?

Edit: Answered my own question when I re-read your post above, nvm!

-Stefan-
Sastrei is offline   Reply With Quote
Old 16th Nov 2009, 07:51 AM   #17
McC
Webmonkey
 
McC's Avatar
 Gallery
Blender Photoshop After Effects
 
Join Date: Dec 2007
Location: MA, USA
Posts: 609
Default

Thanks for the words of encouragement, everyone.

Treybor planted a bug in my head about having an "accurate 3d studio model." Since then, I've been working furiously on this project...just not in Max.

I've been trying to teach myself the underlying math behind photogrammetry. Proper photogrammetry. This has required teaching myself linear algebra. The objective is that rather than obtaining approximately correct estimates for the ship's dimensions, I get 100% accurate ones derived from Surface-from-Motion calculations (i.e. triangulating points precisely based on their image-to-image offset).

There's a lot of complex matrix math that goes into this (camera distortion matrix times the camera transform matrix times the world space coordinate equals the image coordinate times the lambda scale factor ), but the end result should be very worthwhile.

Once I get a handle on it, the first step will be to produce some accurate line-art orthos in Illustrator, which I'll post here for anyone to freely use in making Star Destroyers in the future. Then I'll get back to the actual 3D part. I had several epiphanies working on this over the weekend, including one late last night that might just let me crack it.

Stay tuned.
__________________
Then tells you when or next, than lets you compare. It's easy!

Finished: Ambassador Class | T'Varo BoP
WIP: U.S.S. Coronado | UEDF Illustrious Defender | His Imperial Majesty's...

Ashes of Alour-Tan | Tutorials | Blog
McC is offline   Reply With Quote
Old 16th Nov 2009, 08:10 AM   #18
Meurig
Moderator
 
Meurig's Avatar
 Gallery
LightWave Maya Mudbox ZBrush Paint Shop Pro Photoshop Boujou Digital Fusion Shake Premiere
 
Join Date: Nov 2007
Location: Vancouver, BC
Posts: 4,868
Default

You could also run shots from the film, the higher res the better, through a 3d auto-tracking app like Boujou and if it can get a decent reading from it, it'll generate a point cloud which will give you accurate dimensions quickly and automatically

Same basic photogrammetry theory though.
Meurig is offline   Reply With Quote
Old 16th Nov 2009, 08:18 AM   #19
McC
Webmonkey
 
McC's Avatar
 Gallery
Blender Photoshop After Effects
 
Join Date: Dec 2007
Location: MA, USA
Posts: 609
Default

Quote:
Originally Posted by Meurig View Post
You could also run shots from the film, the higher res the better, through a 3d auto-tracking app like Boujou and if it can get a decent reading from it, it'll generate a point cloud which will give you accurate dimensions quickly and automatically

Same basic photogrammetry theory though.
Yeah, there are several apps out there that do this. I've played with some of them, though the free solutions tend to be hard to use or crash-prone (I'm looking at you, Bundler). There are a ton of research papers on the topic (and now on my iPod...) and I've even go so far as to e-mail a professor at Cambridge asking for some more information.

Ideally, I'd like to create an app explicitly for the purpose of reconstructing reference orthos for modeling from a sequence of images. But, that's a longer-term ambition.
__________________
Then tells you when or next, than lets you compare. It's easy!

Finished: Ambassador Class | T'Varo BoP
WIP: U.S.S. Coronado | UEDF Illustrious Defender | His Imperial Majesty's...

Ashes of Alour-Tan | Tutorials | Blog
McC is offline   Reply With Quote
Old 16th Nov 2009, 08:25 AM   #20
Tenement 01
Registered Member
 
Tenement 01's Avatar
 Gallery
LightWave Photoshop After Effects Premiere
 
Join Date: Dec 2007
Location: Glasgow, Scotland
Posts: 3,557
Default

Are all the Star Destroyer models built to the same proportions in the films?
Tenement 01 is offline   Reply With Quote
Old 16th Nov 2009, 08:32 AM   #21
Meurig
Moderator
 
Meurig's Avatar
 Gallery
LightWave Maya Mudbox ZBrush Paint Shop Pro Photoshop Boujou Digital Fusion Shake Premiere
 
Join Date: Nov 2007
Location: Vancouver, BC
Posts: 4,868
Default

Quote:
Originally Posted by McC View Post
Ideally, I'd like to create an app explicitly for the purpose of reconstructing reference orthos for modeling from a sequence of images. But, that's a longer-term ambition.
We have a proprietary app at work called Photofit which does exactly that
Meurig is offline   Reply With Quote
Old 16th Nov 2009, 08:35 AM   #22
McC
Webmonkey
 
McC's Avatar
 Gallery
Blender Photoshop After Effects
 
Join Date: Dec 2007
Location: MA, USA
Posts: 609
Default

Quote:
Originally Posted by Meurig View Post
We have a proprietary app at work called Photofit which does exactly that
Yes. It's the word "proprietary" that's the sticking point.
__________________
Then tells you when or next, than lets you compare. It's easy!

Finished: Ambassador Class | T'Varo BoP
WIP: U.S.S. Coronado | UEDF Illustrious Defender | His Imperial Majesty's...

Ashes of Alour-Tan | Tutorials | Blog
McC is offline   Reply With Quote
Old 16th Nov 2009, 06:01 PM   #23
McC
Webmonkey
 
McC's Avatar
 Gallery
Blender Photoshop After Effects
 
Join Date: Dec 2007
Location: MA, USA
Posts: 609
Default

Quote:
Originally Posted by Tenement 01 View Post
Are all the Star Destroyer models built to the same proportions in the films?
Sorry, missed this the first time through.

The models from ANH and ESB are different. Devastator in ANH was an Imperator I, which has three heavy turbolaser batteries and a heavy ion cannon battery in each gunwell rather than the four heavy turbolaser batteries exhibited on the Imperator II.

As far as I know, every Star Destroyer in ESB and ROTJ were filmed with the same model, Avenger/Tyrant/et. al. Executor had its own model, of course, as well as several focused models (one for the command tower and its exploding sensor dome, one for the brim trench, etc.).

In addition to the heavy battery change, the other major feature of the Imperator II is the distinctive shape of the command tower's face. In particular, the three "windows" (they're not windows in the proper sense, but they look like window frames) on either side.
__________________
Then tells you when or next, than lets you compare. It's easy!

Finished: Ambassador Class | T'Varo BoP
WIP: U.S.S. Coronado | UEDF Illustrious Defender | His Imperial Majesty's...

Ashes of Alour-Tan | Tutorials | Blog
McC is offline   Reply With Quote
Old 19th Nov 2009, 08:56 AM   #24
McC
Webmonkey
 
McC's Avatar
 Gallery
Blender Photoshop After Effects
 
Join Date: Dec 2007
Location: MA, USA
Posts: 609
Default

Quick update. I finally finished solving the equations for this stuff yesterday (so, that's basically a week learning linear algebra/perspective projection/projective geometry, since it was Treybor's post that touched this whole thing off ). I'm going to try and find some time to write a Python script to solve the equations based on input data tonight and, with any luck, have something to show for all this ridiculous work by the weekend.

If anyone's curious, I'd be happy to share my equations and figuring (y'know, so you can see that I have been actively and relentlessly pursuing this, instead of not working).

I'm also still planning to develop an app that'll accept photos, let you set points (and axes) and then spit out points. Basically, a free (though my fiancee is nudging me to charge a small amount for it), off-line version of Microsoft's Photosynth, or Adobe's (formerly RealViz's) ImageModeler. But, as mentioned before, that's somewhat longer term--I'm a fair hand at Python, but I have done almost no GUI-based programming.

Anyway, hopefully there'll be something nice for you all come the weekend!
__________________
Then tells you when or next, than lets you compare. It's easy!

Finished: Ambassador Class | T'Varo BoP
WIP: U.S.S. Coronado | UEDF Illustrious Defender | His Imperial Majesty's...

Ashes of Alour-Tan | Tutorials | Blog
McC is offline   Reply With Quote
Old 23rd Nov 2009, 09:10 PM   #25
McC
Webmonkey
 
McC's Avatar
 Gallery
Blender Photoshop After Effects
 
Join Date: Dec 2007
Location: MA, USA
Posts: 609
Default

Didn't get any modeling done this weekend, between NaNoWriMo and an ill fiancee. However, I did make some progress in creating an app to do the photogrammetry calculations. Writing it in Python, so it should be fairly cross-platform.

Wish I had something meaningful to show for all the work that's been going into this project lately.
__________________
Then tells you when or next, than lets you compare. It's easy!

Finished: Ambassador Class | T'Varo BoP
WIP: U.S.S. Coronado | UEDF Illustrious Defender | His Imperial Majesty's...

Ashes of Alour-Tan | Tutorials | Blog
McC is offline   Reply With Quote
Old 23rd Nov 2009, 09:15 PM   #26
Mikala
Premiere Member
 
Mikala's Avatar
 Gallery
LightWave Softimage XSI Vue ZBrush Corel Draw Photoshop Combustion Digital Fusion
 
Join Date: Nov 2007
Location: RMH Alberta
Posts: 8,279
Default

Don't worry we know this is a huge project and will take more than a fair bit of time to complete
__________________
PM'S not the smartest place to make veiled threats or derogatory remarks....Just my opinion.
I'm a meatitarian. Ya I know...It's a personal choice. A way of life for me.
Mikala is offline   Reply With Quote
Old 24th Nov 2009, 11:18 AM   #27
McC
Webmonkey
 
McC's Avatar
 Gallery
Blender Photoshop After Effects
 
Join Date: Dec 2007
Location: MA, USA
Posts: 609
Default

Quote:
Originally Posted by Mikala View Post
Don't worry we know this is a huge project and will take more than a fair bit of time to complete
You can say that again!

My fiancee has agreed to get me this when I finish the project.


So, I have some extra incentive beyond the satisfaction of realizing a long-standing ambition. I've already got the smaller version of this kit sitting on my desk at work, but this is the "definitive" version.

Progress on the app will be sluggish for the next week while I buckle down and try to finish my novel for NaNoWriMo (and, y'know, that whole Thanksgiving thing), but I've already got it doing file I/O. It's using simple INI/config-file format for the project files with absolute path reference to source images. I'll probably turn these into binary files later on, but for now it's easier to handle when I can read the data in a text editor.
__________________
Then tells you when or next, than lets you compare. It's easy!

Finished: Ambassador Class | T'Varo BoP
WIP: U.S.S. Coronado | UEDF Illustrious Defender | His Imperial Majesty's...

Ashes of Alour-Tan | Tutorials | Blog
McC is offline   Reply With Quote
Old 24th Nov 2009, 01:15 PM   #28
ChrisGFX
Registered Member
 
ChrisGFX's Avatar
 Gallery
Cinema 4D Photoshop After Effects Premiere
 
Join Date: Aug 2008
Location: Pulheim / Germany
Posts: 272
Default

Quote:
Originally Posted by McC View Post
You can say that again!

My fiancee has agreed to get me this when I finish the project.


So, I have some extra incentive beyond the satisfaction of realizing a long-standing ambition. I've already got the smaller version of this kit sitting on my desk at work, but this is the "definitive" version.

Progress on the app will be sluggish for the next week while I buckle down and try to finish my novel for NaNoWriMo (and, y'know, that whole Thanksgiving thing), but I've already got it doing file I/O. It's using simple INI/config-file format for the project files with absolute path reference to source images. I'll probably turn these into binary files later on, but for now it's easier to handle when I can read the data in a text editor.
OK, lets go!
__________________
If you can’t explain it to a six year old, you don’t understand it yourself.
ChrisGFX is offline   Reply With Quote
Old 1st Dec 2009, 08:30 PM   #29
McC
Webmonkey
 
McC's Avatar
 Gallery
Blender Photoshop After Effects
 
Join Date: Dec 2007
Location: MA, USA
Posts: 609
Default

Right, NaNoWriMo is done (successfully! huzzah!) and Thanksgiving is passed. Time to get back to work.

I made a rather important conceptual breakthrough this evening. I realize I say that every time I update, but it's always true! There's a lot to learn here.

Anyway, I'm actually at the point now where I can't proceed until I get some image data points. I could do this manually in Photoshop, but it'd be way more interesting to actually write the potential to do it into the program, so that's what my efforts are focusing on at the moment.
__________________
Then tells you when or next, than lets you compare. It's easy!

Finished: Ambassador Class | T'Varo BoP
WIP: U.S.S. Coronado | UEDF Illustrious Defender | His Imperial Majesty's...

Ashes of Alour-Tan | Tutorials | Blog
McC is offline   Reply With Quote
Old 16th May 2011, 05:17 PM   #30
3D_CG
Contributing Member
 
3D_CG's Avatar
 Gallery
3DS Max Blender LightWave Terragen ZBrush Gimp Paint Shop Pro Photoshop Digital Fusion Premiere
 
Join Date: Jan 2008
Location: Spokane Washington, North Side in a quiet neighbor hood, Tell I start My Harley, Then its NOISY...
Posts: 4,941
Default

So How Is This Monster Going...

????
__________________

"Remember The Vets who fought for our Freedom" ...
Intel I7 980 OC'D to 5.1 GHZ, Water Cooled, Gigabyte MB
132 gig ram, 5 HD's, TITAN Z Black , GTX 1080,
LightWave 11.6.3, 32 and 64 bit

http://boincstats.com/signature/-1/user/355618/sig.png
3D_CG is offline   Reply With Quote

Go Back   Foundation 3D Forums > Creative Works > Works in Progress
Reply


Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump


All times are GMT -6. The time now is 12:13 AM.


©2006 - 2013 Foundation 3D