Is it possible to create a 3D rendering from a photo image?

[Tony Hunt]

I'm wondering if it's possible to generate a 3D rendering from a good photo that would enable an accurate waterline model to be built from the resulting plans.  Let me explain... 😀, using the two images below as a live example.

 

They show two consecutive views of a small pearling lugger that operated from Darwin in the Northern Territory of Australia, named PETREL.  An interesting little boat, it was built in 1898 in Hong Kong, by the Whampoa Dock Company Ltd, and was then shipped (unrigged) to Darwin as deck cargo on a steamship.  PETREL was registered, so we also have two accurate dimensions to apply to the photo as scale, namely "Length from forepart of stem beneath bowsprit to aft side of head of sternpost = 39 feet"; and "Main breadth to outside of plank = 11 feet".

 

Using these, would it be possibly to draw an approximate plan for a waterline model, then make a 3D rendering that could be overlaid onto the photo so that it is being viewed from exactly the same angle. then adjust the initial plan so that it perfectly matches the photo?  This would then allow all the other visible details to be added in, at which point you would have a reasonably accurate plan for a waterline model.

 

I've tinkered with 2D CAD (without much success!) but I'm amazed at what people can create with the 3D software now available. It seems to open up all sorts of possibilities.  

[Jim Lad]

An interesting pearler, Tony. Very unusual that she was built in Hong Kong, and she was quite small for a pearler. Have you any reason to believe that her hull differed substantially from the usual pearler/ If not, you'd get a pretty accurate waterline model by using one of the few available plans of locally built luggers.

 

John

[Tony Hunt]

She's fairly typical for that period and that location. A lot of the early luggers were remarkably small. A surprising number were built overseas, in both Singapore and Hong Kong, and also in New Zealand.  I guess Hong Kong isn't much further from Darwin than Sydney is!

 

I think you could certainly use some of the plans already in existence as a starting point, then use the photos to refine them. There was a wider range of shapes and styles of pearling luggers than most people appreciate, and the point would be to create a series of small waterline models that illustrated that range of diversity.  Yet another project for the future! 🙄

[Dr PR]

Tony,

 

To answer your original question - no!

 

There are programs that can generate a rough 3D model of something from photos. Multiple images are taken from different positions all around the object. Then with a little prodding the program creates a crude image, often "painting" the original photos onto flat surfaces. I have seen this done to make 3D models of buildings.

 

But I don't know how well this would work for irregular or curved surfaces.

[Kevin-the-lubber]

You might be thinking of lithophanes, where light and shade is sometimes used to confer depth to a photo. Have a look here https://lithophanemaker.com/. Otherwise, I think you could probably get a reasonably true side projection via trigonometry, although I’m struggling to remember how to do that!

[Tony Hunt]

Hmmm. Methinks I haven't explained myself very clearly!  I'll try again.

 

The idea was to create a sketched 3D rendering of the hull, using the photos to get the shape approximately right, then use the clever ability of the 3D software to view the resulting 3D shape from the same viewpoint as the photographer who took the picture and overlay the rendering onto the picture.  Kind of like the beautiful 3D rendering of the the pearling lugger PENGUIN (see below, done by a real Naval Architect, not me!) but much simpler.  Any discrepancies between the rendering and the photo could then be noted and adjusted, until the two aligned perfectly. Yes? No?  It won't work for the underwater shape, of course, but should be possible for everything above the waterline.

 

[Kevin-the-lubber]

Sorry but I have no idea how you’d do that.* If I wasn’t told otherwise, I’d assume the draughtsman for that picture had both the photo and a lines plan or similar, worked up the latter in CAD and married the two together for the picture.

 

* long shot: if the Penguin was 3D scanned, the scan  may have included skin photos and it is relatively straightforward to slice the scan output to get those station lines. But that doesn’t help you with your problem.

[CDR_Ret]

Hi Tony.

 

Here is my two cents (US), for what it is worth. (In this economy, that's not much!)

 

When using photos for such references, it seems that you would need to know at least the focal length (F/L) of the camera lens before even beginning the process you describe. Also, knowing the distance to object factors into this, which determines the angular relationships between objects in the view.

 

Then there is the matter of how these same factors display in the 3D software. For example, Blender allows the user to set the F/L of the viewport and cameras used for rendering the image. You can also set the camera-to-object distance in the program.

 

On the other hand, the perspective view in DELFTship has a fixed F/L that is not stated in the program. (You may be able to communicate with the D/S staff to find out what the program's effective F/L is.) And you have no idea what the viewing distance is to the object.

 

Then there is Sketchup, which I haven't used in many years, that provides both orthogonal and perspective views, but again, the F/L of the perspective view wasn't given (at least back when I was using it). I have no idea if this information is available in the 3D programs like Rhino, Fusion 360, or others used by members in this forum.

 

Another problem to deal with is the foreshortening of objects in the line of sight. Trying to estimate the breadth of the hull at various points along the rail, for example, using a more-or-less broadside view of the vessel would be extremely problematic, if not impossible. The best you could arrive at would be a very rough guess.

 

I've had more than two decades of experience with this kind of frustration using 120-year-old photos of the brigantine Galilee to reconstruct the ship's plans. I have a treasure trove of more than 60 photos of the ship. But trying to conduct photogrammetry to obtain accurate dimensions of various structures has been extremely difficult, for the reasons stated above. Just trying to determine the list of the ship in broadside views turned out to be mostly a very rough guess. Thankfully, I had a reasonable (but not accurate) set of hull plans to start with, from which I was able to make the dimensional adjustments to conform to the photos.

 

Without some sort of baseline dimensions to work with, what you propose would be fruitless, I'm afraid. Perhaps the photographic experts here on this site may be able to weigh in on this topic.

 

Terry

[jmiba]

Hi Tony,

I use the open-source tool Blender for 3D modeling. For example, you can check out my reconstruction of the Green River steamboat Chaperon.

There’s a fantastic plugin for Blender called fSpy that allows you to perform camera matching with your 3D scenes. It’s free to download at https://fspy.io. I’ve had great success using it for my model of the J.M. White.

 

[CCClarke]

I use photos to build 3D objects when detailed plans aren't available.  The process isn't simple or fast, but it's possible if you have the modeling experience.  When I think I'm close, I place the photo in the background and rotate the model to approximate the photo, then change anything that doesn't match up.  These are relatively simple models, which helps.

[Tony Hunt]

Thanks Terry, Jmiba and CCClarke for your replies.  Interesting insights!  I'll give it a crack and see how far I get.

[Tony Hunt]

Jmiba, can you suggest a good tutorial to use to get started 3D modelling ships in Blender?  There are way too many on Youtube to choose from!

[Dr PR]

Tony,

 

You can get useful dimensional information from 2D photos, but you will have to combine all of the 2D information to make the 3D model.

 

First, al the dimensions in a picture are relative to all other dimensions. You may not know the actual dimensions in feet/meters, but you can calculate the relative dimensions of objects. I often work in CAD without defining any units of measurement. Everything is done in "drawing units." After the drawing is done I can set the relationship between drawing units and real world units and the drawing can be in feet, meters, furlongs, parsecs - whatever. So the actual dimensions are not necessary. You can build the entire model from relative dimensions gleaned from photos.

 

Others here have pointed out problems with 2D photos like view angle and distance. If the object in the photo is not exactly perpendicular to the line of sight more distant objects will be relatively small and closer objects proportionately larger. This will be the major problem in determining dimensions of objects in the photo. However, there are a few tricks you can use to overcome this issue.

 

Look for a "standard dimension" in the picture. For example, in most of the 20th century the US Navy had a standard life rail construction on almost all ships. The rails had three vertical bars spaced from the deck 15, 27 and 39 inches, with stanchion height of 42 inches. Since nearly all ships had life rails somewhere, if you have a photo of a US Navy ship from that period you have a built in vertical "ruler" to determine the height of other objects by comparison to the railings. I call this "photoguesstimation." However, the vertical spacing of the stanchions does not seem to be a standard dimension, but varies from place to place.

 

Another trick is to assume that similar structures have similar heights. Deck houses on a ship are usually the same height, or multiples of a common overhead height. Portholes are usually one size throughout the ship, and they can give relative height and length information. Vertical ladders are also vertical rulers. Doors are usually some standard size, although there may be several different sizes on a ship.

 

Remember that the designer and builder used some standard units of measurement (feet and inches, meters and centimeters, etc.). When you calculate a dimension and it comes out an odd fraction of a common unit, just round off to the nearest common unit. For example, I noticed that US warships of the mid 20th century were designed in multiples of 1/16 inch. So I just rounded off all photooguestimatios to the nearest 16th.

 

While you can't feed a 2D picture into a CAD program and have it generate a 3D model, you can derive a lot of information from the picture to allow you to make an approximation. If you do this from multiple 2D photos you may be able to determine dimensions for objects that are common to all the photos.

[That_1Nathan_Guy]

What you are looking to do is properly called photogrammetry.

 

You will need multiple (if not dozens) of photos of the same subject to generate a model that would be anything close to useable.  Even with dozens of photos, there's a good chance you will have to go in and edit the model.

 

Photogrammetry (as well as 3d scanning) generate point clouds, which will usually get turned into a .STL file which is fine for your 3d printer and can be edited, but you will be using something like Blender to do the editing.  Converting it to a parametric file can be done, but it is often labor intensive and will probably require some re-work.

 

Good luck, and have fun in the rabbit holes!

 

Nathan