I have a question about manual alignment mode when merging scans in RevoScan MetroX:
I the feature mode it tries to align the features and does a best fit. What about the manual mode? Does it only align the points I add, or does it take this manual alignment points as a starting point and refines the alignment based on the point cloud like in feature mode? Or, put differently: Do I have to be very precise when placing the alignment points or are they more to guide the software, so a rough “here’s what I want to have there” is enough?
Be as precise as possible. It doesn’t do much beyond the points you set. Unless I’m missing it. It would be nice if I could get it close, then fine tune or maybe something like, align within 5mm so it will try to feature match, but only within a small window. Perhaps it wouldn’t work well. Maybe the ability to manually adjust each part.
Thanks for your answer. So this is from what you experienced I guess? I am asking because a) I think just purely manual alignment I would suppose to turn out terribly as it is not easy to really find the exact spot, b) It turned out surprisingly accurate when I had to do a manual alignment, and c) I think it would be quite “easy” thing to implement, especially as point cloud alignment is what a scanner software has to constantly do anyway
So you experienced it differently, if you slightly misalign the points, the merge will be slightly misaligned, too?
Someone from Revopoint, can you give a statement on that from the developer’s perspective?
Hello, in manual merge mode, the software aligns the models based on the points you select. When selecting points on two models, it is important to be as accurate as possible as placing the points correctly increases the accuracy of the software’s merging process.
If the points you select do not correspond correctly, the models will not align properly.
Thanks a lot for the explanation! However, it is clear that with the example you gave, the software is not able to do a proper alignment. But what, if I would, in your screenshot, on the yellow source, set 1,2 and three somewhat properly, (1 at the wrist, 2 at the tip of the middle finger, three between ring and pinky), but just roughly? Plus-minus 2mm? Then it can lay the two point clouds roughly over each other, resulting in two point clouds with a lot of “near-overlaps”. So it should be easy and great value if it would, starting from this pre-alignment, try to slightly transform the models to maximize the overlap? I still think it is even doing it. Cannot try it out now, but I will later today try deliberately do a manual merge with sloppy point to see if I end up with thick or merged fingers or not.
Maybe an addition to explain what I mean:
If there is no overlap at all, or only very little, of course, then the SW cannot do anything but rely on the precise selection of the alignment points.
If there is a lot of well and unambiguous overlap, no help is needed, the SW can align on its own.
But if there is, for example, a tooth gear with a lot of overlap, but also a lot of ambiguity (which tooth on which one?), or just a model with some overlap, but not enough to unambiguously find the correct match because different alignments would lead to a (local) maximum of overlap, then I would suppose that the SW asks for my help, but then does the final mm alignment itself. Just by finding the next local maximum of overlap around the starting point selected by the user.
Yes, this was also the deliberate effect of carelessly selecting points, just to demonstrate that randomly choosing asymmetric points without accuracy cannot be assisted by the software. If I were to choose more accurately, it could be successfully merged. The more precise the points, the more accurate the merging results.
I saw a Raptor video where the user used marker dots for alignment. I wonder if it would work well for us if the software could detect them and let us pick them for manual alignment. And the point cloud data around markers could help alignment.
I think it does try to use nearby point cloud data. It does not like big holes in the scan. Lol. I got a little lazy and tried it anyway out of curiosity. It did alright, but the iffy data threw it off. When I tried again with a better quality scan, it did much better.
