3D Reconstruction?

Here you can discuss topics such as focus stacking, stitching and other techniques that relate to the processing of micrographs.
Post Reply
Message
Author
ethicalhacker
Posts: 23
Joined: Tue Dec 01, 2020 1:35 am

3D Reconstruction?

#1 Post by ethicalhacker » Fri Jan 01, 2021 11:27 pm

I have a reflected light microscope, and I've noticed that the working distance of up to about a 10X objective is easily enough to allow for significant variation in specimen surface topography with a narrow depth of focus. (e.g. rocks, crystals, circuit boards, jewelry, etc.) I imagine stereo microscopes might be similar (e.g. insects). It would seem like sensors could be rigged up to tell a computer where the stage is in 3D space (might be possible for the software to work around the absence of this with enough calibration) and software could produce a 3D model of the specimen with texture map. A quick Google suggests that this has been done successfully. Does anyone have software or code? I think it would be cool if we could build 3D models of tiny objects by building 3D models of a single orientation at a time and then merging the orientations into a single 3D model.

MicroBob
Posts: 3154
Joined: Sun Dec 25, 2016 9:11 am
Location: Northern Germany

Re: 3D Reconstruction?

#2 Post by MicroBob » Sat Jan 02, 2021 7:25 am

Hi,
much of this is already been done in photo stacking software like PICOLAY. The localisation in three dimensions is determined and you can get a surface map out of the program. What it doesn't do is linking points to lines and forming a 3D volume model. But there are 3D scanning softwares that can create 3D models from a bigger number of photos taken from different angles, I thin kAutodesk used to offer a free software for this. The problem is that you would need photos of every hidden detail to give the software a basis to do the calculation. E.g. youy have photos of an Elk - but do the photos show it's inner ear and every hair from enough directions? Generally this would work best for slick shapes with few crevices, like a jet plane or a single crystal. It would work much worse for objects with a lot of fine detail like found on typical planctol life. And the object has to be persuaded to hold still for how long it takes! :lol:
Usually these scans are not perfect, they need editing. And this needs quite a bit of skill and the right software, Blender perhaps.

Bob

ethicalhacker
Posts: 23
Joined: Tue Dec 01, 2020 1:35 am

Re: 3D Reconstruction?

#3 Post by ethicalhacker » Sat Jan 02, 2021 5:04 pm

MicroBob wrote:
Sat Jan 02, 2021 7:25 am
Much of this is already been done in photo stacking software like PICOLAY. The localization in three dimensions is determined and you can get a surface map out of the program.
I'm still collecting the prerequisite equipment necessarily to experiment with what PICOLAY, etc. can offer.
MicroBob wrote:
Sat Jan 02, 2021 7:25 am
There is 3D scanning software that can create 3D models from a larger number of photos taken from different angles.
A happened to bump into AliceVision Meshroom (GitHub) last night. There are some links to video tutorials on that page that visualize this exact process.
MicroBob wrote:
Sat Jan 02, 2021 7:25 am
Generally this would work best for slick shapes with few crevices like a jet plane or a single crystal.
That makes sense. The first object that actually made me think of this was my [gentleman's] wedding ring. It happens to fit this criteria.
MicroBob wrote:
Sat Jan 02, 2021 7:25 am
Usually these scans are not perfect. They need editing, and this requires quite a bit of skill and the right software - Blender perhaps.
Yes, I am willing to accept this limitation.

MicroBob
Posts: 3154
Joined: Sun Dec 25, 2016 9:11 am
Location: Northern Germany

Re: 3D Reconstruction?

#4 Post by MicroBob » Sat Jan 02, 2021 8:42 pm

This could become a nice number crunching exerise: For a sharp photo front to end you will need about 100 single photos in different focus positions. Then lets say 30 different angles....
To get into the inside engravings you also would need photos from angles that are not so easy to set up. I think VLC viewer can extract photos from a video stream. This might make it easier to the the photos.

Post Reply