Hi there,
I did to 360° scans on the turntable of this sherd 4500 BC and merged the cleaned point clouds afterwards.
Lightning conditions where exactly the same, but you can tell the two scans from the colorrng. Also on some parts the surface is very smooth, on others it is kind of grainy (which I like better besides).
What can I do to avoid this?
Should I not use the turntable to avoid overscanning and shoot “stills” from different angles instead? One scan with pauses and slight angle shiftings?
The same on another sherd, but it’s amazing that it chatched the eyelet in the lug with just 2 scans!
Hi @Archaeoscan.by
You see that the color changed exactly in the areas where the surface have less density than the other , RGB color per vertex relays on vertex density for a proper resolution , how more vertex how sharper the color data will be since each point represents a pixel .
When scanning with MINI it is easy to change distance between scans what can result in uneven scanning density ( and accuracy ) 5 cm can already alter the accuracy greatly .
Normally precision scanning like that should be performed in exact distance for each scanned angle , no scan overlapping , the same settings while capture for all fragments , meaning Manual settings for Depth camera and RGB .
There is a little bug in the RGB settings that change the white balance while the RGB is in manual mode that can cause issues at this moment , but the dev.team know already about that issue .
Please do not over scan it perform “still” shots , the scanner is not build to capture still shots , it will only Overlap the surface , the object needs to rotate at least 30sec per 360 degrees rotation when using MINI
Scanning flatter objects like yours for example can results easy in overlapping , try for example to scan it at a slightly angle meaning put something below the object so it don’t lay flat on the turntable , keep the scanner at 35-45 degrees angle .
For more precise work you can also use Cloud Compare , it have higher level of registration what can improve your results , here is a video I made about how to merge multiple scans in Cloud Compare
I see you uploaded more screenshots while I was posting my reply , I see that your merged fragments have different resolutions what will cause the difference between the sharper and smoother surface . As mentioned before it can be result because if over scanning or change in distance while scanning , the object need to be equalized and remeshed, at this point the RS5 don’t have that function , but the upcoming update will have it already .
Regarding the color change I see it have different white balance of multiple scans , keep the RGB settings on Manual at the same value , however the white balance is not affected by that what is a bug , so the color tone may change .
I don’t know exactly your lighting setup while capturing , but you should have at least 2 light one on top one below the RGB camera for even lighting . And when using MINI not too bright as it can affect the scanning results and induce noises , the LED light colors should be around 4000K to max 5500K , higher than that can add too much blue what will affect the white balance and change tone to warmer color .
it is true that i slightly shifted distance at each scan. So i shouldn’t do that, instead turn depth and color exposure to manual.
But i dindn’t lay the sherd flat on the turntable.
Here you can see my setup, is it ok?
Yes , but the lighting may be better if you put your ring light being the MINI on the left side where is the RGB camera that take the pictures for color data .
I see also discoloration on your object , not sure if it is light related or the color of the object
It is better with MINI to lower the artificial light source down and increase manually the RGB exposure , too much light will affect scanning with MINI especially Depth sensors since it is sensitive to blue light ( Mini Only )
If you can change the ring light tone to slightly warmer than cooler , the results may be better .
Also make sure the light shine at the same angle as MINI …
When you scan Ning taller objects with MINI the top of the object will always have better accuracy than the lower part of the object if the object length is greater than 5 cm … MINI is created for scanning very small and short objects since it range is rather small .
I think you will profit well here with POP3 , as it will give you uniform light in scanning of the color surface and still very good details .
Anyway try to put the ring light straight behind MINI at the same angle slightly in the left side as s center , you should get better color rendition .
Everything starts with a proper point cloud , if it is not good , no matter what you do the mesh will never be perfect .
It is hard to evaluate without actually scanning that object myself … for me looks like too much noises and shiny shading in RS5 that I personally dislike, not makes it looking better .
Please try to do a scan , one 360 rotation , turntable 30sec per rotation and fuse with Advanced mode at 0.08mm and let see the result , without merging multiple angles …
I will do as you told in ther afternoon.
In my opinion, the main problem is that the sherd is to flat. The holes in the mesh are always on the broken edges. So these edges are “underexposed”, while the bigger flat sides are maybe “overscanned”.
I think i could see the mini loosing track slightly sometimes when going from the flat sides over to the edge. It looked like it struggled to readjust.
This might cause this “offset” after merging.
Would it be a solution to put crumpled paper onto the turntable maybe?
By the way, should the extent of the hole object be always in the preview window or is it enough to have the bigger part in the field of view, because the turntable will move all parts into it sooner or later?
I don’t like the shiny shiding in Revoscan, too - do they plan to fix this issue in later updates, maybe with several options?
Thanks a lot and bye until later!
Hi again,
this is one single scan just as you told. Speed at 36 fps, the toggles don’t go further.
Good detail, but many holes and as I tried to merge two PCs, there were seams again.
The surface is also a little grainy, more than in reality.
Regarding color mode, you can see where the piece was scanned twice on the “blurry” impression. I didn’t want the edges not to be cached so, I probably pushed “Pause” a little bit late…
Note that i meshed and textured each single scan before merging for testing purposes. But doesn’t revoscan only use PCs for merging? Overlap MUST have been enough…
you need to clean the overlapped and lose points ( that are no holes ) since you use Standard Fuse mode , try to use Advance Mode at lowest setting I think it is 0.08mm it will clean the pad pints for you .
you need also clean an lose points after meshing in case there was something left
anyway the color result looks very good , if you cleaned the lose points ( no holes ) it would be rather perfect
That is for sure , but you can place another object on the side and scan it together later you can remove it , it would add stabilization for tracking when the object is scanned on the side , try to use Auto exposure under Depth Camera setting .
crumpled paper is good , but with MINI it will add a lot of data and your scan can suffer after fusing on quality , plus you have to remove the paper before meshing or you lose half of the details when scanning without paper .
I don’t like the shiny preview shader in the OpenGL , it is too shinny , they should add more blur to the specular , would be better … hopefully they do in the future as I mentioned it to the team , an options for different shaders would be amazing , and of course We All want Wire Frame preview !
It only uses Fused scan for merging , try to use manual points to find the matching points exactly , I like how the surface looks like today , try to fuse using Advanced mode … just for a test
My idea is that keeping track is they key. I think the two scans have slightly different volumes becaus of struggling to keep track… how to fix this? Slower TT speed?
It should not have different volume , especially after you used support . I think the edge issue is the color , brown absorb blue light so that may create issue but that are just my speculations here .
I like the surface and the single scans looks nice and clean .
Can you sent me the last project zipped to putvmail@gmail.com ? I will check it personally .
Hm…i don’t know - the parts scanned worst are not really the darkest parts. Some spots on the surface have remains of silky polish, but its no metal-like shine.
But if the (brown) color is the problem, POP3 could solve that, right? Infrared light, no blue. I also could try a scan in “dark mode”. Scanning spray i not really an option because the sherds are kind of national heritage and should stay without chemical contamination fot further research.
What i really can’t understand is that the first 2 very neat scans weren’t merged properly, the common features should have bene abundant…
But we will see…
I can send you a download-link for the data tomorrow because i am at work today.
Thank you for your gerat support!
No there should be no issue , you will get also perfect color data , and great details when scanning at 15cm distance what will provide the best accuracy .
that why I want to test the volume myself to find out what is going on to help you out .
OK no problem , take your time and you more than welcome .
Hi Daniel ,
the object merged correctly in RS 5 , the difference you see is becouse the 2 scans have a different accuracy , one has more than the other , both are scanned with different light as well so you cant get the same color results either .
MINI is very sensitive to light changes , so if you scan multiple objects make sure the lighting, and distance is always the same , for the same reason when it get about higher accuracy , the models need to be scanned with 3D sprays to ensure the same accuracy as the color on the surface can influence it as well .
keep the distance and stay away from windows light or cold LED light … I guess POP3 will fix the most issues here with your scans .
As I always said before, you cant actually merge objects from multiple accuracy scans and expect the transition to be smooth . This require already manual edit in 3D editing software.
here the edges between the higher and lower accuracy due to two different distance that 2 parts was scanned at
Thanks a lot.
It is true that the toggles for fusing accuracy switch without my intention sometimes because they are controlled by the mouse wheel now. sometimes. I made my own tests in the meantime and was more or less sucessfull. Waiting for pop3 now…
