Submicron photogrammetry is possible!


Submicron photogrammetry.

I would like to share my latest 3d scan of something very small and familiar to everyone, yet tricky to scan as it is quite shiny metal.
The Goal in my research was to acquire the most detailed 3d scan with as low noise as possible to extract displacement maps from it.
As well as ultra detailed diffuse color maps for texturing.

Thanks to ultra precise motion controlled focus stacking rail I was able to get all 7200 images stacked pixel perfect.
For reconstruction I have used reality capture with alignment error of:
minimum 0.2pixel and maximum 0.5 pixel.

for capturing images i have used canon 7d. This is not the best camera for focus stacking as it has quite low resolution and it’s a dslr. with a mirrorless camera process could be much quicker.

On stacked images one pixel corresponds to 410nm which surprisingly allows 3d reconstruction to capture recognisable features under 1 micron.
For reference, the visible part of the ball in this pen is approx. 660um
There is some signal noise which makes it harder to notice features under a micron, because I’m not sure what is signal noise and what is a 1 micron real feature.
Things get easier when you have machining marks on this pen or scratches which are traveling over long distances and it is relatively easy to spot by rotating light in a 3d package.

I’m sure I haven’t hit the limit yet as I know I’m able to do it better than this, but I just don’t really see the reason why. I almost reached the possible limit of what visible light is able to deliver.
(i know i could go to ultraviolet but this would mean no more color textures)

This would conclude my over a decade long chase for the most detailed and textured 3d scan possible.
So I can finally close this chapter.

veni, vidi, vici.