Quote Originally Posted by 24c View Post
.. I also would like to access the scans as is, before they are meshed, so maybe then I could combine them in Meshlab. I did save some simple closed models on the turntable, by doing one pass in two different positions, and then aligning these completed scan files externally. It'd be nice to have an .obj export, but as Nerv says you can use Meshlab for that too.
Hughes did you get any further saving the point cloud scans (.asc??). I've seen the raw scan file as they are date and time stamped in the EinScan-S program files, but I've no idea how it works. presume they start at 0 and go to n number of scans increasing by 1.It looks like the software saves an edited version of the scan, and an alignment file, and then this is incrementally added too?? ..
Hey Mike,

I did have to use Meshlab in the past few days as i was scanning the side of my bike and there was no way i could do a 360 scan. The meshed file of Einscan was ok but as i chose low detail initially some parts were too approximate (Note to Alfred: would be nice to have the option of meshing the scan in higher resolution after we saved it first in low res. I think you can do this because i assume your point cloud files are the same regardless of the resolution we choose). So coming back to Meshlab, here's my workflow, and it can surely be improved, comments welcomed:

- In Einscan software, at the end of your scan, save the file in ascii format instead of stl. It will create one file per scan with the XYZ coordinates of each scanned point. These files are already aligned to each other
- Start new project in Meshlab - Import all point cloud files, keep checkbox marked, then click ok for each file
- Select first file in layers window, right click, select Flatten Visible Layers, check last checkbox Keep... This will group all your scans into one file or layer.
- Make it visible on screen by clicking on its point cloud icon in the layer window, or else you don't see it
- Align camera to approximate position of scanner (i'm not sure if this is necessary, this is to help the system create normals)
- From the menu: Filters/Points Set/Compute Normals for Points set: Select 100 neighbours and check Flip Normals...Also enter 0, 0, 1000 in coord, then Apply . This step is necessary to tell the system which face of the scan is actually facing the scanner, or else, they are just points without any notion of top or bottom. This is where i'm struggling. At the end of this operation, i could still see some black regions on the scan, probably means some normals are inverted and this will be annoying for the next step
- Then Filters/Points Set / Surface Reconstruction Poisson: use default for a firts rough check, then redo with Octree Depth set at 10, and 8/1/1 for other params . This gives a pretty good detailed mesh. The only issue for me was with the normals calculated from the previous step.
- Then you can export your mesh in stl format, or other format too, i think obj, but i did not try.