Reconstruction of dimensions fom a photo…

[Robin GrahamParticipant @robingraham42208]

Here is a pic of an embossing tool  I made for a local ceramicist:

I made several wheels, but only one holder.  He now wants another holder but can’t give me the original to measure because “it’s in constant use”. And of course I can’t find my construction notes if ever I made any.

What I’m wondering is if there is any way of reconstructing the dimensions of the holder from the photo when given:

a) the diameter of the wheel

b) the information that the ‘corners’ of the holder are at 90 degrees.

I sort of think that mathematically it could be possible because (b) might allow calculation of the perspective angle from which a ‘face on’ perspective could be generated and then (a) would give a metric.

I realise that the most direct approach would be to tell the guy that if he wants an exact copy he’d better give me the original to measure or no deal, but now I’ve got interested in whether it’s possible to do this sort of reconstruction.

Robin.

 

[Robin Graham]

On 26 August 2026 at 02:30 Robin Graham Said:

…can’t give me the original to measure because “it’s in constant use”.

Make life easy for yourself and call him out for the liar that he is.

There are 168 hours in a week. It is not going to be in his hand every one of those hours.

If it were a teenager and a mobile phone, the constant use statement might be plausible.

Tell your customer to lay it on a sheet of toilet paper while he is otherwise occupied in that room and take a photograph of it. The distance between the perforations will serve to scale the rest of it. A good customer would send you in the post a sheet of said paper (unused) for verification.

[JasonBModerator @jasonb]

To scale dimensions off an image you will need at least one known measurement from which you can proportion all the others. If the pen knife is yours then use that or the guy must be able to take a couple of measurements for you such as wheel dia and the AF of the square stock.

When you have a know size post again and I’ll tell you how to do it

[JasonBModerator @jasonb]

Or you could look at your previous post from which the diameter f the wheel if made to the same size can be deduced from the roller and steel rule.

A workholding question

[John HaineParticipant @johnhaine32865]

If he’s local pop round and measure it.

[Robin Graham]

On 26 August 2026 at 02:30 Robin Graham Said:

He now wants another holder but can’t give me the original to measure because “it’s in constant use”.

 

As Mrs Malaprop might have put it:

With fiends like that, who needs enemas !

MichaelG.

[Robin GrahamParticipant @robingraham42208]

Thanks for replies.  The ‘fiend’ (nice one Michael!) has left a spare wheel with me, so I know its diameter.  What I was wondering was if it is possible to reconstruct the dimensions of the 3D  object given the diameter of the wheel and the fact that the ‘corners’ are 90 degrees.  It has the character of what maths people call ‘an inverse problem’.  They can be difficult, but having seen what modern 3D CAD can do with the ‘direct’ problem (ie generating arbitrary 2D projections given a 3D dataset)  I wondered if the inverse process had been cracked.  I suppose someone with CAD skills could draw the general shape and adjust dimensions/projection angle until it fitted. But true inversion may be an ill-posed problem.

I shall probably get my hands on the original, one way or another, and make the thing though the ‘customer’ can sometimes be frustrating!  But my purpose here was really to enquire about the feasibility of the reconstruction of a 3D image from a single 2D projection given the constraints I’ve indicated.

Regards, Robin

 

[Michael GilliganParticipant @michaelgilligan61133]

Hypothetically, Robin … Yes, it is plausible

In practice, however the lens distortions and/or the camera’s inbuilt corrections thereof, make it very difficult indeed.

The ‘uncertainties’ when attempting to the vanishing-points have always defeated me.

MichaelG.

[Michael GilliganParticipant @michaelgilligan61133]

This is a good introduction to the principles … somewhat complicated by his mention of two Apps]
.

.

I have Morpholio Trace, which is excellent in many respects … but I struggle to do this stuff with the requisite accuracy.

MichaelG.

[Wade BeattyParticipant @wadebeatty78296]

<p style=”text-align: left;”>Some modeling programs (I use Rhino 3D) allow you to import a photo and use it as a background that you can then scale and use as reference. As Michael pointed out, distortion is always an issue, is the photo completely perpendicular and centered?. I have had success and failure with the process.</p>
<p style=”text-align: left;”>I have hired a photometrology camera set-up to measure equipment subsea with a remote submarine and the results were a fantastic and accurate 3D model, cost was beyond belief but critical enough to be worth it.</p>

[JasonBModerator @jasonb]

Post what the diameter of the wheel is and I’ll show you what is possible. I do it all the time with old engravings or photos of engines to pull a buuch of sizes off to enable me to design a new model. Probably 1″ if the photos with a steel rule in your other thread are anything to go by.

There are some programs that can take into account perspective or you can even construct your own vanishing points on a CAD drawing, the two red lines on this one go to a vanishing point

[JasonBModerator @jasonb]

Round up the decimals to nominal imperial if that is what you work in and you should be off to a good start. Easy enough to alter if teh wheels were not 1″

[egaParticipant @ega]

Did anyone mention asking AI to do it?

The problem demonstrates the value of working to standard dimensions.

[peter1972Participant @peter1972]

Being rather pedantic, it would be better to allow for components further away appearing smaller. Here the two green lines are intended to indicate how much smaller, in proportion, the flywheel appears than the cylinder because it is further away: about 7%.

[JasonBModerator @jasonb]

Luckily I have several image sof those pumps and can take an average or use the best view for any particulat item. based on keeping the fins the same diameter as they are at eye level my 215 flywheel on this drawing is not far off 7% increase on 200mm. 😊

I mainly just picked that image as it is recent and showed a vanishing point

[Robin GrahamParticipant @robingraham42208]

Thanks very much for your reconstruction Jason, really helpful.  The wheels are actually 26.7mm diameter, guess the customer asked for ‘around 1 inch’ and I arrived at that to get an integral number of ‘dimples’ at the right depth/spacing. Obviously I can scale from your pic, but the holder dimensions aren’t that critical so probably don’t even need to do that.

Regards, Robin.

 

[JasonBModerator @jasonb]

[Robin GrahamParticipant @robingraham42208]

That’s bordering on the uncanny Jason.  The central hole in the wheel was almost certainly  finished with a 5mm reamer. 10mm for the OD of the bronze bush also fits with the  materials/tools I had at the time.  Thank you again. Can you reveal how you did it?

Robin.

[JasonBModerator @jasonb]

I made a copy of your image and then imported that into Alibre.

I then sketched two lines over one of the wheels and show the length they measured. As you can see these were quite a bit bigger than the known dimension.

The dialogue box for the image allows me to scale the image in X and Y. I have done X and you can see my entry in Y as 26.7 / 71 which Alibre will convert to a decimal as it has done with X. It can also be done by using the callibrate scale but I have always done it as described particularly if the image is squashed up in one axis as yours is.

With that done any further lines that I draw will be to the correct scale. It does need a bit of intuition for example my line may have been 4.92 for the hole but as you gave the OD in metric I suspected you work in metric so 5mm was the obvious number to go with.

The other options in the box can be used to position my 0,0 datum at any point by moving the image up, down or sideways and I can also rotate it if the image is not level. You can see I have moved this one to put the axis through the central pivot

I can then start taking off dimensions knowing whatever line I draw it’s length wil be to scale and in most cases below from a known datum point

[Michael GilliganParticipant @michaelgilligan61133]

Nicely demonstrated, Jason

… a MasterClass to start the day

MichaelG.

[Robin GrahamParticipant @robingraham42208]

Thanks for the detailed explanation of your procedure Jason – I really appreciate your taking the time to do this.  I now have all the data I need to reconstruct the holder.  Maybe someday I’ll get round to learning CAD!

Robin.

[Author]

Posts