Photogrammetry

[JimG]

Big word.

But I'm thinking of trying to use it to sort out the dimensions of the Caledonian 498 Class locomotive which I'm starting to build. There is no official GA drawing available from any of the drawing archives, but there is a good range of pictures available. There is one, small scale drawing available on Jim Smellie's Caley Coaches web site at the end of the notes for his 4mm etched kit, but I'm not sure that I agree with some of the dimensions on that drawing after pouring over a lot of the photographs over the past few weeks.

So over the past day or two I've been digging around on the web to see if there is anything available which would allow me to do something using the computer and I've come across a lot of information about photogrammetry which uses two or more images to generate three dimensional information about an object. I've investigated a few but they require that the pictures be taken at the same time in a specific manner and on the same camera.

I'm thinking of having a look at this software to see if it will work with images from different cameras, and also in black and white.

http://www.arc-team.homelinux.com/arcteam/ppt.php

I'm also aware of AutoDesk's 123D Catch

http://www.123dapp.com/catch

but have been put off by their EULA which gives them full rights over any images used and I don't have copyright on any of the images I have.

has anyone got any experience in this field so that I don't go re-inventing the wheel, or wasting my time.

Jim.

 

[ZiderHead]

I've done this in 2D with my own shots using a short tele, shooting 10MP RAW square-on to the subjects, then adjusting for perspective (tilt/shift) in SilkyPix and it turned out to be pretty accurate.

Trying the same with scanned and rather blurry old colour negs was considerably less accurate, I imagine creating a 3D model from old images would be pretty marginal in terms of getting useful measurements out of it.

That said, if there are no blueprints or measurements available it may be the only way to go. The key will be getting the very best possible scans from the source images (negs, slides or plates preferably).

Jon

 

[SteveO]

I'm not entirely sure this technique will enable you to plot an etch, for example, although it will produce an approximation of a 3d image that you can rotate and investigate. I used the basis of a similar technique called stereophotogrammetry to plot the model to produce etches for my Fox Walker, but I only used the most basic reference plots as I was doing it manually and in Adobe Illustrator – and I only had three photographs. I think I used about 40-50 reference points (very very low!) to build the basic shape then flattened it which gave me the essential patterns to plot onto the etch. But even that took a couple weeks work.

 

[Steph Dale]

Big word.
But I'm thinking of trying to use it to sort out the dimensions of the Caledonian 498 Class locomotive which I'm starting to build. There is no official GA drawing available from any of the drawing archives, but there is a good range of pictures available. There is one, small scale drawing available on Jim Smellie's Caley Coaches web site at the end of the notes for his 4mm etched kit, but I'm not sure that I agree with some of the dimensions on that drawing after pouring over a lot of the photographs over the past few weeks.

So over the past day or two I've been digging around on the web to see if there is anything available which would allow me to do something using the computer and I've come across a lot of information about photogrammetry which uses two or more images to generate three dimensional information about an object. I've investigated a few but they require that the pictures be taken at the same time in a specific manner and on the same camera.

I'm thinking of having a look at this software to see if it will work with images from different cameras, and also in black and white.

http://www.arc-team.homelinux.com/arcteam/ppt.php

I'm also aware of AutoDesk's 123D Catch

http://www.123dapp.com/catch

but have been put off by their EULA which gives them full rights over any images used and I don't have copyright on any of the images I have.

has anyone got any experience in this field so that I don't go re-inventing the wheel, or wasting my time.

Jim.

Jim,
Photoshop Elements has a 'perspective' tool which I've used to flatten pictures. It actually works best from a shallow 3/4 view as then lens effects don't have a significant effect. In any case I wouldn't expect straight lines to come or all that straight unless the images has been taken on a camera with an exceptional lens.
I'm not sure you're heading in quite the right direction with this, you may find that just flattening some images and manipulating to known dimensions will be enough. And quick.
Steph

 

[JimG]

I've done this in 2D with my own shots using a short tele, shooting 10MP RAW square-on to the subjects, then adjusting for perspective (tilt/shift) in SilkyPix and it turned out to be pretty accurate.

Jon,

I've tried a bit of messing around in Gimp with the perspective tool with some limited success, but I was never really convinced that I was getting something that I could trust.

Jim.

 

[Dog Star]

Jim,
Photoshop Elements has a 'perspective' tool which I've used to flatten pictures.

This facility is useful for "unscrabbling" the sides of PO wagons... and has revealed some liveries which were unknown previously. The technique is explained with examples in an issue of the "Railway Archive"... using a photo of Wantage yard as an example and thereby introducing us to a wagon from the Wantage Engineering Company (cue historians for chapter and verse on the Wantage Traction Engines... ).

Sopwith (Pup), we shall be watching for your skills in writing the sheeting of a corresponding model.

regards, Barking who is looking for the RA reference...

 

[JimG]

I'm not entirely sure this technique will enable you to plot an etch, for example, although it will produce an approximation of a 3d image that you can rotate and investigate. I used the basis of a similar technique called stereophotogrammetry to plot the model to produce etches for my Fox Walker, but I only used the most basic reference plots as I was doing it manually and in Adobe Illustrator – and I only had three photographs. I think I used about 40-50 reference points (very very low!) to build the basic shape then flattened it which gave me the essential patterns to plot onto the etch. But even that took a couple weeks work.

Steve,

I wouldn't be looking to generate etches from the work since I would be scratchbuilding the loco so I'm only really looking for positional accuracy for key features of the loco. I did have a look at some of the web based stereophotogrammetry sites but they seemed to require pictures taken in a particular way and with the technical features of the camera known. I did try to use a couple of the old pictures but they were rejected since there was no camera information in the files - presumably this information being a key part of their calculations.

Jim.

 

[JimG]

Jim,
Photoshop Elements has a 'perspective' tool which I've used to flatten pictures. It actually works best from a shallow 3/4 view as then lens effects don't have a significant effect. In any case I wouldn't expect straight lines to come or all that straight unless the images has been taken on a camera with an exceptional lens.
I'm not sure you're heading in quite the right direction with this, you may find that just flattening some images and manipulating to known dimensions will be enough. And quick.
Steph

Steph,

As I said to Jon, I've tried using Gimp's perspective tools but I haven't been totally convinced that the results were accurate. I'll maybe have another go and see if I get anything more convincing.

I can see where photogrammetry might help if I drew up a basic framework for two or more pictures - say lines along the footplate edge (the known dimension), lines up the front and rear tank edge, lines at front and back of smokebox, etc., etc. and then do a large amount of number crunching with some yet to be found software. I dare say that one advantage I might have is that most taking lenses up until the middle of last century tended to be of a focal length to match the human eye - e.g. around 50mm for the 35mm frame and about 80 mm for the 2 1/4" frame, so most of the pictures I have - 1912 to 1960 - will probably have similar lens characteristics.

Jim.

 

[ZiderHead]

Jon,

I've tried a bit of messing around in Gimp with the perspective tool with some limited success, but I was never really convinced that I was getting something that I could trust.

Jim.

Hi Jim

It is subjective using perspective tools in Gimp, PS etc on scans or images without EXIF data as you are basically doing it by eye, and it really depends on having decent references in the image to work off.

The major advantage of using your own digital images is that they have the EXIF information embedded, and this can be used by a perspective/distortion correction plugin (PTLens is really good) so its calibrated for each individual image.

Of course this doesnt help much with historical images unless, as you say, you know the focal length and can enter the values manually.

Jon

 

[Steph Dale]

Jim,

I haven't tried Gimp, but do know that the perspective tool in Elements works well, it's helped me produce workable illustrations for the carriages I've built.

Your assumption on focal lengths seems reasonable but may well be flawed. Any photos taken on rangefinder or viewfinder cameras are likely to be semi-wide angle. 35, 28 or even 23mm would seem more likely on a camera up until SLRs came along in good numbers.

Not a show-stopper from your requirement, but certainly something to be aware of.

Steph (who has a number of the sorts of cameras I'm referring to.)

 

[JimG]

I had another play with Gimp this afternoon with questionable results again. I think the problem I have is that Gimp does work best when the object to be adjusted is a plane surface and that is a bit difficult on a locomotive - especially a Caledonian tank loco with cab sides inset from the tank sides, and a large radius curve to the top of the tanks, which makes life difficult trying to assess the tank top accurately. I'll give the matter more thought. I was hoping to settle the frame dimensions so that I might have started on cutting the frames, but I do need to settle the cylinder placing before I can contemplate that.

I've also sent off to the NRM for a GA of the Caledonian 782 class tank. This was one of the most numerous locomotive classes on the Caledonian and was the prototype for the Wills 4mm white metal kit of many years ago. But it was also the fore-runner of my dock tank since the dock tank was reckoned to be a squeezed version of the 782 class with smaller wheels and a shorter wheelbase. That might allow me to use the smokebox placement and dimensions from the 782 Class GA which would them let me set the cylinder position.

But I'm still intrigued with the possibility of using photogrammetry. Many thanks for all the input on the subject.

Jim.

 

[JimG]

I couldn't put it down. I did a bit more searching around on the internet and came up with this.

The second tutorial which follows on is very interesting since it shows the use of a second photograph to check out work on the first. It might be worth a bit of messing around to see what the results might be.

Jim.

 

[Dog Star]

Your recent post has rung the bell.... our late Guv'nor had experimented with Sketchup to sort out the size and placement of buildings from photographs. I cannot recall the details suffice to say that Sketchup could work with aerial photos to produce ground plans, Guv'nor described this in several posts to this forum (probably around the time when the Wallingford diorama first came to light).

Hopefully one of the FS can pin-point the relevant posts.

regards, Graham

 

[JimG]

Your recent post has rung the bell.... our late Guv'nor had experimented with Sketchup to sort out the size and placement of buildings from photographs. I cannot recall the details suffice to say that Sketchup could work with aerial photos to produce ground plans, Guv'nor described this in several posts to this forum (probably around the time when the Wallingford diorama first came to light).

Hopefully one of the FS can pin-point the relevant posts.

Graham,

Maybe this thread?

http://www.westernthunder.co.uk/index.php?threads/sketchup.765/

Jim.

 

[Dog Star]

Yes that thread seems to be what I recall. How time flies, that discussion was two years ago and just as relevant today; our late Guv'nor had a habit of teaching us "how to" rather than saying "today is Friday, have some fish".

regards, Graham

 

[Renovater]

Post removed

 

[adrian]

I just clicked on the Thread to see what that word meant.....

and are you any the wiser for having done so?

 

[40126]

I just clicked on the Thread to see what that word meant.....


...staying at the Back, hiding under my desk....

Which word ?, Gimp ??

Steve

 

[SteveO]

My first post here, so Hi. I have tried another method which i do trust and can certainly help to get as close as possible to the dimensions needed or at the very least put you in the right direction. If you have a basic dimension such as axle distance then you can do the following. first blow up the image to A4 size, then draw a line from the top left corner of the locomotive to the bottom right corner, then the bottom left corner to the top right, where the lines cross is the centre of the locomotive. from the centre draw a vertical line straight up, this forms a new box and so do the same process over again, in the end you end up with multiple boxes getting smaller and smaller. Now take the axle distance dimension and relate that to the the vertical lines of the boxes, that will give you the dimension of each box, from there you can calculate the positions of all, IE cab length etc, easier done than said.

I think I understand but can you post an example picture please? I'm stuck on the drawing the boxes bit as they'll all be the same size and not drawn with perspective.

 

[Renovater]

Post removed