[Blackpool] The Make:Shift:Do project - photogrammetry,

[M Hewitt]

For the Craft Council Make:Shift:Do project at the end of the month, we
want to build a system which can scan a component/item and produce a 3D
file which can be fed into the 3D printer to manufacture an identical
component/item.
here is an example: http://freelss.org/

Anyone with an interest in this, please get in contact/involved.

Here is a wikipedia article with lots of photogrammetry software, 24 of
which are free.
https://en.wikipedia.org/wiki/Comparison_of_photogrammetry_software.

Of the 24, 3 are web based.
Web based systems do not need us to have software or a high powered
computer.
The trade off is the time and bandwidth required to upload the required
number of high resolution pictures, and the complete lack of control over
the workflow.

here is one example:
http://www.arc3d.be/     Free web powered converter.

Here is the 3DF Zephyr  (not web based)
https://www.3dflow.net/technology/documents/3df-zephyr-tutorials/convert-photos-3d-models-3df-zephyr/
The generally suggested hardware specification to run such software is:

i7, 8G min, 16G preferred memory and a CUDA capable graphics card.
Depending on the software, CUDA is not always required, but slows things
down if not available.

I am sure several of us have have capable hardware to test the various
software, and Tony has kindly offered to loan a suitable machine to the
club.
I would be interested in knowing how long such a machine would take to do
the job.
Quote from ARC3D cloud based system mention above:
"Depending on the size, number and quality of the images that have been
uploaded, a typical job may take from 15 minutes to 2 or 3 hours. "

On the picture taking side, a lot of writeups are using the pi camera on
it's own or in multiples, but
equally,  some are suggesting better results are obtained with higher
quality pictures.

I have seen everything from a smartphone to a high end SLR suggested and it
seems like one camera is good enough to get the job done. Multiple cameras
do not seem to be required. Several members probably have high quality
cameras.

In summary,  we have everything we need to start testing options.

*******
Software.
Some articles describe pipeline style workflows, where defects/holes in the
mesh can be identified, and extra pictures of the area can be taken and
added to correct the problem.

But maybe we will be looking for "pictures in, 3D printer file out" with no
intervention from the user.

Here is a good article describing the different workflows.
https://pfalkingham.wordpress.com/2016/09/14/trying-all-the-free-photogrammetry/

Written by Dr. P.L. Falkingham who wrote this white paper in 2012:
Acquisition of high resolution three-dimensional models using free,
open-source, photogrammetric software
http://palaeo-electronica.org/content/issue1-2012technical-articles/92-3d-photogrammetry

Dr. Falkingham says this about Agisoft Photoscan, one of the two
suggestions Olly picked out in his first email.
"This program has become something of a standard among colleagues who use
photogrammetry, and for good reason.  At $59 for the educational standard
version, it’s a bargain, and it’s easy to use interface means anyone can
use it. "
Is there anything in the rules to prevent us buying software if it turned
out to be the best option?


This is an open source pi lazer scanning kit.  Similar in concept to what
we want to attempt.
http://store.murobo.com/atlas-3d-kit/   $229

About FreeLSS
FreeLSS is a free as in open source, open hardware, and open electronic
design 3D printable turn table laser scanning platform based on the
Raspberry Pi. It is written in C++ and licensed under the GPL. The scanning
software runs self-contained on the Raspberry Pi without the need for a
connected computer via USB. The user interface is completely web based and
is exposed via libmicrohttpd on the Pi. Laser sensing is performed via the
official 5 MP Raspberry Pi camera. The camera can be operated in either
video or still mode. Video mode camera access is provided by the Raspicam
library. Reference designs for the electronics to control the lasers and
turn table are available as Fritzing files. Access to the GPIO pins are
provided by wiringPi.
Features

    Fully 3D Printable
    Point cloud export
    Triangle mesh export
    Assisted calibration
    Support for dual laser lines (right and left)
    Up to 6400 samples per table revolution (with reference electronics)
    5 megapixel camera sensor
    Support for camera Still mode and Video code
    Configurable Image Processing Settings
    Ability to generate images at different stages of the image processing
pipeline for debugging
    Persistant storage of previous scans
    Manual control of lasers and turn table
    Flexible architecture

Formats
FreeLSS can generate results in the following formats.

    PLY - Colored Point Cloud
    XYZ - Comma Delimited 3D Point Cloud
    STL - 3D Triangle Mesh



Mike