I am working on a metal belt buckle which has lettering and symbols in relief on it (elevation of various physical features). The belt has a metal surface, with a dull matt metallic look. Upon photographing I had no problem with getting the detail sharp without any distortions. I got nice photo’s without shiny spots (Probably because of the matt dull metallic surface structure). So my guess is that the surface of the object is good enough for RC (without having to spray the object with matt paint like the “shiny object tutorial”).
My photo base in alignment is good enough, aligning about 400 photos. I have done a preview reconstruction to see what the alignment will produce. My next step is probably reconstruct in high detail.
My experience with previous projects is that I can’t seem to get the surface of the object the same as in real life with reconstruction. I have not done a reconstruction in normal or high detail yet, but to go ahead of things I am wondering how I will get the best detailed reconstruction?
In my previous project I had a problem with the reconstruct, because smooth flat surfaces ended up grainy in RC. One of the reaction pointed out that RC has difficulties capturing shiny- or one single color objects.
What can I do with the reconstruction settings to upgrade the appearance of the object? To make the surface resembling the real thing?
In conclusion, what can I do to copy the exact object in 3D with the same structure/smoothness of the object’s surface qualified for 3D printing?
I had exactly the same question with my last scan. I did not check jet but here are my suggestions i made, will test it in the next scan.
When i take many 24mp closeup Pictures, what level of Detail is enough. My idea is to rescale images outside of RC and then always reconstruct in high detail to check wat pixel density is enough to reconstruct certain details in a good quality. I think 400 Closeup pictures at 24mp are far to much.
Checking what difference it makes to Change “Detector sensitivity” for flat surfaces. I think the grainy look comes also from over detection of features
What really helped to capture reflective surfaces was to use a polarized filter. Also to take Photos with a huge overlap resulted in nearly flat surfaces. Also when your surface is dull, there is reflection.
In my case i made reconstruction at normal with downscale 2 (so 12 MP) and one time at high, (24MP). The only result was i had more detail, but not significant less grainy surfaces.
Surface noise can have various different reasons. Problems with the alignment are quite likely. Especially if you have a difficult material and, as Oliver has already pointed out, perhaps too many images. The problem then is that neighboring images may have too small a basis (relation of distance from each other to the distance to the object. Another possibility is noise in the images.
RC works in ways that make it almost impossible to get perfect results with flawed image sets. You just cannot save it by tweaking the settings. The only feasible way is to learn to take the images properly. Sounds frustrating if you are a beginner but I learned that the hard way…
Hey, checked arround a bit and made a scan of an RC Car Model Part, the result is impresive. I coated the hole part with water color to ensure to have great detailed surface. Then i made about 156 fotos at 24mp
So my conclusion is -> the texture of an object is what you need to get a good surface finish right after the scan.
Also my second “turntable” scan, i put the object just onto a white paper and turned it by hand. I used also 2 Softboxes from ebay.
Also checked what difference it makes to use a polarised filter, it removes some dull reflections from the surface.
Source Image:
Made one top, one bottom 360 turn. To ensure alignment, i also made a 360 turn form side view.
The result(4,1mio tris), at 12Mp (Normal detail)
Will also try a second run at high detail, just for fun
I think this scan needs a minimal amount of post processing, i think some bool operations to open up the closed holes.
Is there any good way to set the scale right? i dont understand what to do in realtiyCapture.
My idea is to 3d print this, so would be nice to know how to set the scale.
hello, there are more ways how to scale scenes in RealityCapture - have a look at the Scale the Scene tutorial in the Help section directly in RealityCapture
Yes, after playing around, reading the manual several times, i managed to set the correct scene scale. I was a bit confused because of the new options and a message about no ground control plane set. So my workflow was now (mixed up with the new handy point suggest tool)
After reconstruction process, click on “Control Points” and place 2 points in the 3D view
Then click on “Suggest Points”: RC suggests images for this point
Go through the list of images for each point, when the suggest is correct click onto the green apply button. You need at least 2 images for each point.
Click on “Create Distance” and connect the 2 Points in the 3D view, (or enter the point names manually)
Set the “measured distance” value (default unit setup is meters)
Click on Update.
Export your mesh.
To set the points in 3D view is much more accurate than to do it manually in 2D view, but also depends on the object.
