Posting 3-D models

3-D model created using Scaniverse on iPhone 13 Pro. and shared as .fbx file to be viewed on 3d Viewer or similar. (.fbx files can be created with viruses embedded but mine are safe to use)

This is an experiment to see if I can paste such a model onto this forum so no promises of success.
This post may come and go as I try to get it to work.

Success! If you click on the file link it should open a new window and say " Cannot open this file type " or something similar. Click on the download box (below that message) and the downloaded file can then be opened in your 'fbx supporting 3D viewing program.

Is this just being curious, of is it part of a project? This type of scanned file is rather difficult to use for CAM for example. I've being playing with a scan of the PR tailpiece to see if it's possible to 3D print a filler section, to lift the rear up a couple of cm, so that I can get a better match between the (Indian copy) petrol tank and the seat unit.

SteveBorland said:

Is this just being curious, of is it part of a project? This type of scanned file is rather difficult to use for CAM for example. I've being playing with a scan of the PR tailpiece to see if it's possible to 3D print a filler section, to lift the rear up a couple of cm, so that I can get a better match between the (Indian copy) petrol tank and the seat unit.

Click to expand...

At the moment I am just playing and exploring what can be done for little or no cost. It seemed to me that a 3D model can replace a video or several photos. The ultimate aim is to create models of small parts which can be 3d printed. I don't know if that is feasible. They almost certainly need post processing for that purpose because they are shells and need some thickness or to be solid, possibly with some correction/refinement of the mesh. I haven't explored available, cheap, software for doing that yet. Proper CAD software, such as solidworks, is rather expensive. Phones with lidar, like mine, create quite accurate accurate models.

I've done quite a lot of work with scanned files, using high end optical systems such as the GOM ATOS (https://www.gom.com/en/products/3d-scanning/atos-compact-scan). Even though they are industry leaders (and wit ha price to match!) the output is not really suitable for transferring to a CAM system to replicate it in metal. The mesh is very large, very complex and requires an awful lot of work to clean it up. Then there is the problem of finding hole centers with a sufficient degree of precision. Some of this work can now be done automatically in software, but it's a very long way from being a good CAM mesh.
It's a bit difficult to find any hard numbers to describe the accuracy of these phone based scans, but in general, the larger the scan is, the worse the precision will be. The software is trying to do a best fit with all the images it records, and since there is no reference framework to align each photo, there will be an accumulation of errors as the scan is built up. For example, if you start scanning a tyre, by the time you have done a complete 360 degrees, the first and last meshes will be at a considerable distance from each other. The software then tries to compensate by forcing a fit between the first and last image, but what about the
rest of the scan?
The only way to solve this is to use a reference grid, usually a pattern of randomly placed dots on the object to be scanned. You then take a lot of photos of this from different angles and thus build up a 3D map of the frame. Then you use this to align the individual scans .... it's a lot of work and needs high precision, but it gives very good results. it's not a home solution though :-(

That doesn't sound very promising. Perhaps I would be better off just making interesting scans to post on the forum.
When I used my last phone to scan a cylinder head the start and finish didn't line up just as you report but, using the newer one with lidar the start and finish did seem to line up. I also noticed on the old one that a rough cast surface (just like your pattern of randomly placed dots) scans much better than a polished surface where there are few reference points and it gets confused by the reflection. When I scanned the inside of the ports, drawing a random grid on the surface worked quite well.
Back to the drawing board!

Norton 850 Roadster


Its a scale model but the tank has a decent shape but there is work needed to add the front fitting notches.


Plan to scale up and print on my 300x300x320 3d printer

hybridracer said:

That doesn't sound very promising. Perhaps I would be better off just making interesting scans to post on the forum.
When I used my last phone to scan a cylinder head the start and finish didn't line up just as you report but, using the newer one with lidar the start and finish did seem to line up. I also noticed on the old one that a rough cast surface (just like your pattern of randomly placed dots) scans much better than a polished surface where there are few reference points and it gets confused by the reflection. When I scanned the inside of the ports, drawing a random grid on the surface worked quite well.
Back to the drawing board!

Click to expand...

Yes, the rough surface will scan better 'cos it's not reflective. The best scanning surface is light coloured matt - you can buy a scanning spray which puts a very thin coat of white powder on a polished or transparent surface.
The reference grid I mentioned is actually produced in separate process, using photogrammetric software.
For something like the Commando head, I would be looking for an accuracy of about +- 0.05 or better.
For small objects, we used to scan parts for mobile phones down to about 10 microns, so the results are very good indeed.

The main use of the system though is to compare a physical part to the original CAD model, to check the accuracy of the part. For example, all cars today use flowing curved surfaces, which are impossible to measure with traditional methods, so optical measurement is the only way to go, using a difference map with colours indicating the divergence from the original CAD file.