Here’s what I had - I am currently working on PC so will show the point cloud and send to you later. But it is essentially now a blob.
I did delete overlapping data (in individual scans and after merging). However, this deleted way more data after the fact. I am certain that was not what happened. Meshing should not be impacting point cloud data imo.
I treat point cloud like original data and avoid making edits to it just in case I want to redo something later -except cleaning up data (overlaps and deleting background points). However when meshing or closing the program, my data disappeared. It’s odd if you’ve never experienced this.
When I get a chance, I’ll try to replicate it a d send you the file.
I should add I was using a POP2 for the challenge, unlike many that use Range.
This is probably what happened but I was under the impression, from your previous post, meshing is independent of point cloud.
As an improvement, I would highly recommend not editing point cloud when meshing. PC is hard to find overlapping or miss aligned scans. Meshing helps identify problem areas imo.
Maybe create a secondary PC if needed to make edits during meshing process.
Basically, you should be able to go back and forth between mesh and point cloud, deleting data, adding scans, etc. Without losing resolution. This is not intuitive and other software that I’ve used doesn’t do this - there’s no need to continuously compress the point cloud everytime you mesh.
Yes I experienced that couple of times already , after meshing , saving the project , the mesh was not available after opening again the project and I was back to the Fused point cloud again , what is certain a bug .
It should not, but it did also in RS4 , I believe the mesh is converted to point cloud after meshing if this is not the original RAW project data and that is very odd since each time you mesh , less details will be available .
Ok great , that would be useful to find the bug by replicating the issue with the same scan data .
The challenge was for Range users in general , it is more difficult to scan this types of object with POP2 but the basic you scanned looking not bad at all .
Yes it was a challenge haha lots of scanning and rescanning as I had to use padding often to merge scans but then had to delete padding and rescan as this caused padding to become overscanned. I also had to use a combination of marker scans for flat areas and body scans for larger FOV.
Disappointed I was not able to complete the challenge but It shows that POP2 is capable of scanning larger items like this with a lot of patience.
I will send you better screenshots & data later when I am able to do so.
It should not happen, the original point cloud data needs to be intact … I will investigate it closely again and see if the latest RS5.0.6 have this bug again .
I just trying to find out what happened in your case , as that is not the desired result , this issue was in RS4 … but so far I did not experienced it in RS5 yet , possible POP2 profile have a bug , I will check it out and let you know .
Below is the merged scan after meshing (probably more than once - we see the point cloud is now very uniform and very much less dense than the underlying scans.
Ignoring the problematic areas (specifically the legs) - we see that details of the spindles, padding trim, and hand carved decorations on the back are washed out and no longer have the detail of the meshes posted above.
@PUTV I believe you are right and that when meshing the algorithm edits the Point Cloud. I am unsure if the RAW data is still there as unfortunately I deleted the individual scans as I worked to keep things organized. Perhaps the RAW data is untouched when meshing - but the fused point cloud gets compressed/edited. This is still undesirable IMO.
There are no issues with Scanner (Calibration shows .13XX or better point resolution) and below are scans of before/after attempting scanning furniture. The artifacts are just cause by bad merges or overscanned areas.
You mean that is point cloud is already fused right ? ( No mesh ) mesh and point cloud are 2 different data files.
You see the fused point cloud appears dense because of the multiple overlapping , the hardware can only produce a point distance that is listed in your device specification , and over scanning it will not produce more dense point cloud as this is technically impossible , to check the proper density of your scanner after meshing , you can import the mesh to CloudCompare software and measure the distances between the 2 points , what in case of Fused point cloud from POP2 is 0.05mm to 0.3mm
Range between 0.1mm to 0.4mm
Mini 0.02mm to 0.2 mm
It also depends of the distance that the objects was scanned at as this affect the pitch point resolution as well .
I tested RS5 optimal meshing results and was equal to CloudCompare in RS5 .
I will test tonight your project that you sent me and make some POP2 tests to see if the meshing vs fusing altering the main point cloud as this looks like happening again … but I will be more sure about after the test .
You see , all you really need to capture a single object at one single 360 degrees rotation is 12 frames , not 400 or 4K , everything else what was captured is partially already removed while fusing to create the correct pitch point distance according to your scanner … you can’t add anything more to the point cloud as the additional data is just lose junk that helping tracking while scanning but not really nesesery for the final point cloud , maybe 20% to max 50% depends of overlapping of the frame cells and speed while scanning .
Yes, the raw data was fused and overlapping points were deleted and PC isolated deletion were performed using RS5 point editing options like shown in YouTube videos. Neither Point Cloud Smoothing or Simplification were used. On the first image, no mesh was created from the single scan, instead it was merged with the rest of the data and meshed in the full scan.
Basically what I’m trying to say is yes I agree with you the density will decrease when omitting bad data - though - the bad data should have been deleted using isolation & overlapping deletions (both were performed). The remaining data should pretty much stay the same - though it does not. There are ~116k points in the scan segment and ~215k points in the merged scan which the merged scan should have 4x more points instead of ~2x in my opinion as the size of scanned area is much larger than the segment. The scan density (standard accuracy) was used for the entire project and when merging new scans you can see a big difference in densities from the point could data to the existing data. Unfortunately I am unable to demonstrate that currently as the existing data won’t merge with the scan shown probably because the mesh is not dense enough.
I need to get CloudCompare but haven’t used it yet.
You forgetting about the all overlapped scans areas while merging , that need to be removed . Normally after scanning I get 5 milion points , after cleaning 2 milion points then you have to remove all overlapped parts with the other scans what will result in much smaller point cloud at the end , almost 70% smaller than the begining , on top you can’t mesh the full 100% point cloud , you still will lose points in result no matter what software you are using .
That why I always suggest to scan one angle rotation of an object at 360 degrees to preserve as much data possible for easy cleaning and merging .
Regardless, I think the bug is present. It still reduces the point cloud beyond what it should probably due to meshing more than once as you stated.
If you look at my first image of an early mesh in my process and the last mesh sent to you the resolution is cut significantly.
Again, my belief is that the fused mesh should not be altered during meshing and doing so makes scanning and checking the scan quality more difficult. If the mesh needs to be altered during meshing as it’s currently done, then the software should make a temporary duplicate to mesh - leaving the original data intact.
In GOM, you can go back and forth between mesh and point cloud and edit either and the Point cloud remains intact during the meshing process.
On small scans this is not an issue but when merging a lot of scans, it is a tedious workflow.
That’s the way it should work in RS5 .
It is possible the old bug slipped in with the last version again, I will investigate it tonight .
If the bug is present, it will alter small of big or merged scans .
Scanning service is one of most expensive one for a good reason, lots of work and time is involved , especially when dealing with partial scans . However merging big scans require lots of RAM , and 16GB is good for a single scans in general , since some of my bigger scan take over 32GB in size what would never be possible to process with a system operating on 16GB of RAM .
However I have your project and will send you back the results soon for evaluation.
i was working on the furniture challenge also, but i gave up when I finish the mesh and process the texture, no iidea why it spent 3 hours and no processing, progress bar around 15% for 3 hours…no cpu process and ram not using much
RevoScan5
Please tell me the start timer function from the older RevoScan is going to be implemented in version 5. Also it would be nice if we could move the small Rendering window on the screen. Sometime my scans get lost behind this window. Maybe even have it pop up or open after pressing Complete on the scan.
im using the latest version of the firmware, and wondering if its true , Range cant be used in sunny day, coz I find Range cant capture anything when I scanning outdoor during sunny day…
