Trying out the Inspire, and I need to scan parts of a wall, door and floor to 3d print a fancy fitting piece for skirting board with a size of abt 3mm x 10mm x 800mm. The scan was successful, but shows the floor as having a “bend” of a few millimeters, something it does not have in reality. Now I am wondering what would be the best method of either preventing this “non-straightness” in the scan, or fixing it somehow. Is there a way of defining known reference points along the scan, something that could be used to make sure scan stays flat from start to end? Or some better solutions?
either use markers or add features for the scanner to align to.
crumpled paper or crepe tape can work wornders
I already have those in place, or actually I used a towel :). The issue is not that the scanner would not track, the scan is perfect in that sense. Everything is nicely scanned, no holes and very little noise. The only issue is that the surface that gets created is not true to life, it deviates by a few mm. As I did the scan by moving the scanner by hand, it feels like it might not be a wonder that the surface does not align with extreme accuracy. And now I would like to understand whether there could something more that I could do to “calibrate” the scanned surface. I believe the issue would be somewhat reduced if I could scan the other end of the surface, and merge the items together, but with a floor that gets complicated 
adding some pictures of your capture/ issues would help
1 Like
Seems I can only add one picture, as I am a new user. So here is the scan viewed from the left side, which shows that there is a few mm dip roughly at the center of the door. Does that help? I can try adding a photo on another post, if that might be allowed.
Here is the same door, where I placed a piece of skirting to show that in real world the dip is not there.
I’m sorry it’s not clear to me where is the added features on your scan.
where is the towel?
seems there is little data for the alignment to grab on to. have you tried using marker mode and the magic mat?
I have not. The scan had no issues staying in sync, and scanned very easily with everything (needed) becoming “green” as I progressed the scanner along its path. The surfaces to be scanned are not totally featureless by themselves either, so I probably could have done the scan even without the towel, but since it is such a easy thing to add I use it wherever I can. I did not see any indication that something would not have worked, or that there would have been a problem with anything. And unfortunately now I cannot retry, as the target is off-site.
But regardless, from a theoretical point of view, the scan will always have some error, and with single sided surfaces even small errors can accumulate, as the scanner does not have an accurate information about its location or a reference point it needs to align to. It does have accelerometers, which can be used to estimate the (relative) position, but those will also have errors, and the positional accuracy will degrade over time, the longer the scan lasts. I do not know the details, but I would assume that the main method for creating the surface and the distances is indeed stitching/aligning the scanned frames together. And that process will inevitably have errors, since the point cloud frames are not absolutely accurate either, and will have noise and potentially other errors that cannot be averaged out. Thus I am not surprised to see non-zero accuracy.
However, judging from your comment this level of error is unexpected, and I should expect to see (more) accurate results. What would be your expectation of the errors seen in this kind of a scan? Does Revopoint provide estimations that would be relevant here?
unless you have zoomed in on that dip you shown and that value is below 0.5mm between the top and bottom, I am fairly certain that your scan just did not had enough features for propper alignment and therefore twisted.
i reccomend you try with the magic mat and marker mode to rule out bad registration