photogrammetry - how to get complete solid model

I'm just exploring this area and have a fundamental question that I'm sure others have solved.

As an example I have a badly corroded flange which is a part from a boat and the part is not commercially available.
I want to create a 'master' for molding, so have scanned the object and edited to have a visually correct 3D object in DesignSpark PCB, Autocads 123 and Meshmixer.

Here I'm stuck! The part has a hole where it was sitting during to the photography, so now realise I need to repeat it in a better fashion to get all of the 3D data and will repeat by balancing it on three spikes (which I can subsequently delete, but how do you fill the remaining holes?


Other questions/comments:

It's interesting to me on this first experence in this area in that you obviously create a surface and in any hole you can visualise the inside of the object. Fascinating, but how do you go to the next step of printing a replica? Does the object print as a solid or is this also another step in the process?

Also having scanned the object is there a method of editing such that the corrosion pits and scratches are removed/filled? Obviously I do not want a perfect replica! I require to produce something similar to the original manufacture.

---

Ormerod #007 (shaken but not stirred!)

For something relatively simple such as a flange I would use my CAD program of choice, measure the original item and create a 'real' CAD model.
For more complex geometries I would probably try the scan and edit process that you are currently using.

---

Bob Morrison
Wörth am Rhein, Germany
"Luke, use the source!"
BLOG - PHOTOS - Thingiverse

Thanks Bob,

Agree totally and that is what I have done in DesignSpark Mechanical (which I will be using as a comparison), but for this exercise I would like to see what can be done in photogrammetry and associated tools.

So questions still remain how to fill the missing part/holes etc?

---

Ormerod #007 (shaken but not stirred!)

There's a video from 123DCatch that shows how to take your photos for good quality photogrammetry:

youtube - 123DCatch

It talks about 123DCatch, but I think the tips apply to photogrammetry in general. I've gotten some nice solid models following them. Though I haven't yet turned them into STLs. You can correct defects in the software.

One thing I don't think it mentions is that you want to use the highest fstop you can - at least 8. You need good depth of field, because the process uses details from the background as well as your subject.

A couple of my models:

Krishna
Buddha

I've spent around thirty hours to make this 3D model out of 28 photos.

Software I used are VisualSFM, Meshlab.
(https://youtu.be/V4iBb_j6k_g)



Suppose I can take many directional photos about an object with great resolution, creating a 3D model for 3D Printer or other application seem to depend on computing power.

I think I now have a mobile 3D scanner.
- DSLR camera with a tripod
- Post processing software
- Debian Linux workstation

For small object needs a flat surface and a camera provided decent PC that crunches image data.

Edited 1 time(s). Last edit at 06/01/2015 09:32AM by janpenguin.