Hi,
I tryed the Fabscan Pi for the first time after building and I got very poor quality in the scan results. The bottom part of the object is not acquired, and the same for the left side of the object.
I calibrate the scanner but at present I cannot 3D print the stand for the calibration sheet and I created one with cardboard. Even if the calibration ends successfully, I noticed that the camera picture cuts off a small part of the scan plate on the left (but the camera and the led ring stick very well to the wood wall).
Any idea?
Note: I will post the results of two scan of the same object (a "no color" one with default resolution and threshold, and a "color" one with max resolution and threshold) when I will be able to find how to attach pictures to this post.
Thanks and best regards,
Iade Gesso, PhD
PS: happy Easter :)
Have you tried to adjust your origin point in config. You will find additional information on how to do that in the documentation.
Hi Mario,
I followed your suggestion and I chacked my origin point. In the default configuration it was 37.2 but I lowered to 33 since i don't have the calibration stand and the calibration sheet touches directly the scan plate... but no right model is acquired.
Here the result: https://ibb.co/7jBShHz
Regards,
Iade Gesso, PhD
What does the point cloud look like? There is also an issue with the meshing option since the upgrade to 0.7.0. Generally FabScanPi is using meshlab under the hood. You can try to load your point cloud to meshlab and mesh it by hand until the meshing bug is fixed.
1.) Filters -> Point Set -> Calculate Normals
2.) Filters -> Samples -> Poisson Disk Sampling ( choose a value, I use 2/3 points of the source cloud. e.g. if the point cloud has 100.000 use 60.000 points)
3.) Filters -> Remeshing, Simplification and Reconstruction -> Screened Poisson Surface Reconstruction
This will give you a good starting point. But play around with the parameters. Watch Mr. P Meshlab Tutorials on youtube this will help a lot.
(maybe start with this one)
At the end the most important thing is to get as much points and density to your point cloud as possible. You can also try to do multiple scans of the object in different object orientations. You also can use the tutorials to find a solution for merging point clouds. Another powerful tool for this job is cloud compare.
Hi @mario,
here the point cloud of the model I show you yesterday: https://ibb.co/BPfV5MX
While waiting to have time to see the tutorial, I started setting the resolution of the camera module to its maximum values (I have a Pi Camera v2). I forgot to tell that to reach the results you can see I had to focus the camera manually.
Regards,
Iade Gesso, PhD
Hi @mario,
I tryed to download the point cloud to use it directly in my local copy of Meshlab (I'm on a Gentoo system, connected to the Fabscan through my wired gigabit LAN), but the .ply I downloaded seems to be without the EOF marker.
Moreover, changing camera resolution hangs on scan start.
I tryed to better center the camera to get less background, but never changed.
Regards,
Iade Gesso, PhD
