Bump… still looking for some guidance on this.
For anyone else who’s following along, here’s some visuals. In the middle, we’ve got a point cloud that shows all the merged scans. Around the outside, we have the individual scanned pieces that were merged to create it.
(In case anyone was curious, it’s the trunk of a car. I’m scanning it so that I can model up replacement steel parts that will just drop in rather than needing a ton of hand-fitting.)
This first corner is the front left. All the flat surfaces appear flat, and the corners appear to be 90 degrees… but they’re not.
Here’s the second corner- front right. Still looks pretty good. We’ve got plenty of scan data, there’s a little bit of noise, but you can see that most of the stuff that looks like noise is the same in all the scans. It’s rust, built up dirt, bubbling paint, etc. We’ve got plenty of overlap in the scans, and it looks like they merged just fine.
Again, the corners look like they’re 90 degrees, but they’re not… and that’s adding up.
Third corner. Right rear. Great detail- you can clearly see the texture on the strips of sound absorbent rubber that were glued to the floor panels. The sides of the hex nuts on the hardware look good. There’s even some places where I can see spot welds in the scan that are difficult to see in person with the naked eye. The detail is great.
The corners look nice and crisp… but again, they’re bigger than they’re supposed to be, and that error is adding up.
And here we are now at corner 4… and corner 5.
Those errors in the previous corners added up, and this is where we came out. All four sides of the trunk should be largely parallel to each other, and when merged, the two corners that you see here should only be one corner.
Here’s a top view with an orthographic projection. You can clearly see here that things which are supposed to be straight are very much not straight, and that angles which should be square are not.