With the latest version I’m getting better and better results with point cloud fusing.
Till now I did mutiple scans to seperate ply clouds and later fused them in other software.
I started to trust more and more the workflow in Revoscan:
scan for a couple of hundered frames
fuse pointcloud
scan some more - let Revo automatically align new data to old cloud fuse
fuse again
the new fuse is often surprising as it deals with some misalignments you see during scanning - you can see dual layers or ghosting but after fusing the model is perfectly clean - great job guys
This workflow is actually getting, dare I say it, reliable
There is one thing that still stops me from using it all the time.
If a new scan/fuse is not great - we get dual layers, ghosting etc after fusing - game over - we have to start from 0.
So this is my suggestion:
every time Revoscan makes a new fuse of new scan data and an already existing point cloud fuse, please save the previous ply as a separate file - with a suffic _L0.
So let’s say we go through the above process 5 times, so we do 5 consecutive scans that each time get attached to the growing fused point cloud - we should end up with our
project_fused.ply file
but also
project_fused_L0.ply,
project_fused_L1.ply
project_fused_L2.ply
project_fused_L3.ply
This way, if the last scanning/fusing fails and we get a corrupted pointcloud after final scan, we can still revert to L3 file that was the point cloud before the last fusing operation.
This way we loose only the last scan.
Why over complicating everything
you have export button for each fused point cloud , do that , and merge all your point cloud data in Revo Studio or other software of your choice as all professionals do in this world .
Each time you start new scan the old fused points get unfused even after it is meshed already, since it is converted back to unfused point cloud data .
I would prefer a software controlled turntable that scan each angle at 30 degree , fuse it and go for the next angle for a perfect result , I did it manually yesterday , the results was so much sharper , 55 frames and full scanned 360 rotation with excellent details and accuracy .
PUTV, this is where I am at. I have the DIY Slider working, and it’s strength is how much control over the; turntable, scanner height, and scanner angle it allows.
I have played with this “Snapshot” every 30 degrees also.
And as kaczorefx noted the ability to regain tracking after fusing, is magical. Seems to only work if I fuse the scan first, before rotating, or flipping the item.
That what I was doing yesterday , fused the point cloud before moving 30 degree and worked like a charm, no tracking issues, no overlapped point cloud frames . My turntable has only 45 degree automatic rotation , but I used the 1 degree button pushing it manually 30 times , stopped fused and next , time consuming but the result was great.
I wish I could do that , sadly I am completely green when get about the stuff you are building , just in theories , now if I can turn my turntable each 30 degree in place of the standard 45, that would be already perfect . I can still build a manual one lol always better than pushing 30 times the remote control
I thought I was actually simplifying it.
The point fusing is getting really good in Revoscan - why would I do it in a separate software?
I will clean it up later, but the fusing alone is usefull.
Another advantage is scanning consecutive scans in context - you can see what you have scanned before and fill the holes - it gets complicated if you treat each scan separately,
I know it’s kindergarten simple if we’re scanning on a turntable, but unfortunately scanning small objects that can be fit or brought to the turntable is just a small part of the 3D scanning industry,
Even you would benefit from that - you have all the intermediate scans in separate ply files, you can fuse them yourself in MeshLab or CC if you want. You can even automate the process using the Revo final fused pointcloud as a guide for rough point cloud orientation - no need to pick point on separate point clouds - you’ll get your scans orientations automatically, all you have to do is iterate to reduce delta errors, which is also automatic both in MeshLab and CLoudCompare.
Sorry, I see what you mean by that now .
However I was not talking about fusing point clouds frames in a different software as you can’t export unfused point cloud nowhere . It only works in Revo Scan .
However importing fused point cloud for merging anywhere else , it unfuse it again anyway.
What we talking here is not fusing , but alignment of the point cloud frames that is happening inside Revo Scan exclusively , and can’t be done outside Revo Scan . The name for the button is Fuse , but it is actually alignment of the single point cloud frames you captured into new clean point cloud data that could be saved automatic until next step as you suggested saving time .
Yes.
I think it includes aligning and some cleanup operations, as after fusing the resulting point cloud is much better than during scanning, but yes. I meant whatever is hiding underneath the Fuse button
It gets me thinking if a indexer could somehow be used for doing this. Maybe having a stepper motor turn the hand crank. The benefit is you could put a lot of weight on one of these indexers.
You can mark each 30 degrees on the lazy Susan , or buy yourself phone operated turntable for less . Problem is that most electric turntable with remote have only 1 degree or 45 degree and never 30 degree and that is the maximum you can go , anything more will lose the tracking quick , my older green laser 3D scan , capture stuff in 9 shots at 40 degree and has automatic turntable via Bluetooth. I was thinking to reuse the turntable for POP2 but 40 degrees may be too much for the Revo Scan smooth tracking .
Should we ask Revopoint for a snapshot feature?
Instead of a sequence of measurements, just a single one each time we press the button.
Not exactly sure how depth measurements are made, but maybe giving it more time for a single frame would even result in better quality data - similar to lower noise in image sensors with longer exposures.
That would be great , but if you want all the functions you should buy Einscan SE , I saw it sold by Revo before with a different name , wondering if Revo sold it to them .
I think the goal here was to make POP2 portable 3D scanner , plus MINI going to use the same software so all functions and settings need to be neutral .
I made yesterday some interesting testings adjusting manually the Infrared Light structure reducing the noises almost to zero at 15-20 cm distance without using denoiser. The type of technology POP2 is using will always create noises to some degree , how greater the distance how bigger the noises and less details what is a normal behavior .
You remember my Athena sculpture ? This type of material produced the most noises free scan because it was slightly IF absorbing like human skin is . For the same reasons 3D scanning spray produces specific layer reducing the scattering of the IF reducing the noises . The white color really don’t matters here , what really matters is the surface .
Right now the main settings of the IF light are much too strong to scanning white slightly reflective objects at 15-20 cm distance, even for the included bust . Any amount of red you see in your Depth cam preview will produce noises on the surface , so what I did was reducing it manually by creating specific IF filter I placed over the IF projector and it worked . I could scan white objects at 15-20 cm distance and mesh it at max 6 without noises or using denoiser .
Holy crap, thank you.
I was so baffled why white objects give me more noise than even transparent plastic (dunked in flour ;))
If they can’t reduce the sensitivity of the IR sensor they should have some ND filters for it.
Come to think of it I have some strong ND filters for my DJI Air 2S, they are around the right size - if they work, I’ll print a casing for the pop2 with a slide in slot for the ND filters
You need a strong ND filter at least 32 , IR photo filter would be better as you can reduce the level by rotation , but I used my ND gel filters to check if it works, first I tried to reduce the sensor but I lost the sharpens then I used it on top of the IR and boom , my white object stopped glowing red and perfect noiseles scan .
However now when using different mode I have better control over it . The gain setting at 1 is too high for Feature mode , you need to be at least 20cm away when scanning white objects .
Filter ON, meshed at 6 , no denoiser . Feature mode, distance at 15cm
there will be still little grain effect as it need maybe little stronger filter but not like before , plus when I set denoiser ON, it is smooth completely , for that result I used before denoiser at 10, now at zero.
couple of markers and here you go
