How does Shape Fusion work?

[Austin]

I put the shoe last on a turntable and make two scans,the turntable rotates 20 degree.
And I choose the Around Y-Axis, Rotation Angle is 20.
But can't get the right fusion result. How can I get the precise result?
Thanks
===
Austin

[WalterMo]

There are different ways to align a series of scans which were scanned on a turntable. The following example with an antique Opel should show how I mainly do it.
The most important requirement is that the axis of the turntable is the very same as the coordinate system used by DAVID. But also 'Alignment around' x and z can be used if you orientate the turntable axis horizontally parallel to the corner walls. See the sketch.

The car was mounted on the vertical (Y) axis and was scanned at 8 positions à 45°. The direction of rotation was counter-clockwise (top view). If you choose clockwise, you have to set the 'Rotation Angle' in DAVID as a negative value.
I have only used the first three scan positions for this example. The files are named as:
Opel2 0° (green), Opel2 45° (blue) and Opel2 90° (ochre). After uploading I have clicked on the button "Arrange visible scans in one row". The vertical line below each car is the axis of the turntable to carry the car. Often it's better not to put the object on the rotating disk but to extend the turntable axis and fix it there. See screen shot 1.

Starting with alignment:
Choose the 'Alignment' 'Around Y-Axis', set the 'Rotation Angle' = 45°.
Click on 'Align Scans'. Click on the first file 'Opel2 0°.obj', then click on the second file name 'Opel 2 45°.obj'. And voila, the first two scans are aligned, see the second screen shot (2).
Now I prefer to combine both to a "group":
Click with the left mouse button onto the first two file names by pressing the 'Shift' key at the same time so that the names are red shaded. Then click with the right mouse button on this red field, a new small window opens. Click with the left mouse button on "Combine selected scans". Result:
See screen shot 3.
Now set the 'Rotation Angle' to 90°. Click on 'Align Scans' and in the 'List of Scans' on '[Group]'.
Then click on 'Opel2 90°.obj'. See the result (after some seconds) on screen shot 4.

This was shown for only 3 positions. Continue your work in the same way, combine each time to a (bigger growing) group and after all positions are aligned, delete all unwanted things (like here the car holder) and then fuse them.

Hope all was explained correctly.

Walter

Edit: The final alignment after 'Alignment around an Axis' and the preset angle should be made by the tool "Pairwise Fine Registration". Maybe several times to achieve the best result. X56GHY.

[Austin]

Thanks Walter.
There are some questions about your example.
"The car was mounted on the vertical (Y) axis and was scanned at 8 positions à 45°. Opel2 0° (green), Opel2 45° (blue) and Opel2 90° (ochre)."
After your first Alignment, Opel2 0° rotates 45° and aligns to Opel2 45°.Then you combine them as a group.
The [Group] should rotate 45° and align to the Opel2 90°.Why do you set the 'Rotation Angle' to 90° not 45 °?
Or you should first select Opel2 45° and then Opel2 0°(because the 'Scan A' is the moved one) .And Opel2 90° then the [Group]?

[WalterMo]

Today I have talked with Sven and he suggested not to combine scans in a growing group. The reason: Finer alignment because the overlapping areas can better be used.

I had done it in this way and then you are right Austin, for each new alignment of the next scan the constant angle of 45° must be set.
In my case -45° because my turntable was turned counter-clockwise.

Screen shot A shows the 3 single scans:
Opel2 0°.obj (green)
Opel2 45°.obj (blue)
Opel2 90°.obj (ochre)

First alignment:
Scan A = Opel2 45°.obj
Scan B=Opel2 0°.obj
See screen shot B.

Second alignment:
Scan A = Opel2 90°.obj
Scan B= Opel2 45°.obj
See screen shot C.

After all alignments are done the tool "Global Fine Registration" should be used. See screen shot D with the 8 single scans.

[Austin]

@WalterMo,@Sven
I've followed the steps Walter posted to scan a shoe last.
But after the "Global Fine Registration"(3 times), the scan result is not so good. Indeed ,it is deformed.
How to scan the objects like the shoe lasts which have not so much distinct shape?
And how to get a much more precise result? I've tried some scans and the scans cut by the machine is more than 0.2% precision of the whole one.
Thank you very much for your help!

[hyppyayo]

Austin I think we do a similar job, me also in shoes!

[benoy]

Austin finally were you able to get a good fusion of the various scans?
Benoy

[benoy]

Waltermo,

In this mail you have mentioned that

The most important requirement is that the axis of the turntable is the very same as the coordinate system used by DAVID. But also 'Alignment around' x and z can be used if you orientate the turntable axis horizontally parallel to the corner walls.

An easy way to get the Y axis parallel to the rotation axis of the turn table is to place the calib panel on the rotation table and then calibrate. Am I right?

Again if the entire setup is rotated the (That is the rotary disc of the turn-table with the Calib panel on top of it). the Y axis remains the same . Am I right ?
Thanks

Benoy

[WalterMo]

Dr. Benoy,

If the disc diameter is big enough to carry the corner and the corner was built of exactly rectangular walls it should work.

Walter

PS. Often I use my turntable without the disc. The motor axis is extended by a rod and the object is mounted at its tip. So it's sometimes easier to decrease the distance object to the corner walls. Of course this is only true if I scan with the background walls.
Another advantage not to use the disc is that the object is better accessible for the camera and the laser / projector.