Synchro mode to take advantage of the linked turntable

Hi all. We now have the MINI equipped with the multi-axis turntable. But the best feature of the turntable is that it is linked to, and controlled by the scanner. This means that the scanner is aware of the rate of rotation of the turntable, and of the angle of tilt at any given moment.
So why isn’t this information being used? Whenever the turntable tilts at the start of the new revolution, the scanner goes haywire; it often loses alignment and produces misaligned fragments of the point cloud, or even loses track altogether. The experts advise us to only do one rotation at a time, then stitch the result in the editing program. This makes no sense—the turntable was supposed to help, not hinder scanning.
I propose that a new scanning mode be introduced, besides the existing Feature and Marker mode: the Synchro mode, where the scanning software takes advantage of the information it has on the speed and position of the turntable to correct the misalignments creeping into the plain Feature mode. The scanner KNOWS fairly accurately how the position of the model changed since the last frame; it knows, because its settings determine how the turntable has moved since last frame. This information can help align points.
Granted, the accuracy of the turntable’s movement is not on the same order as the spacing of points. But if I understand correctly, even approximate awareness of the actual movement of the model can narrow down feature search, and introduce a “sanity check” into the algorithm. The Synchro mode has the potential to make the linked turntable an unique strength of the Revopoint scanners, rather than a messy complication.

The scanner is not aware of anything , not the rotation or amount of the frames or the angle, it don’t control the turntable or is in synchronization with it . As before it only track the features of the model no matters it is regular or dual axis Turntable.

The start scanning button just starts the turntable rotation sequence set up by you and nothing else .

Simple command, one button to start both devices at once.

But it would be nice to make the software on pc aware of what the turntable is going to do and compile things accordingly.

Of course that would be nice , I would be happy already if it stop scanning while changing the position as it add a lot of not nesesery frames , wasting time removing it later.
That is the difference between 21K vs 1K 3D scanner software.

But there is still little room for improvement, hopefully sooner than later but lower your expectations, you will be less disappointed :wink:

At this moment I seriously prefer my YT-1000 dual axis turntable, as I have greater control over , can use it with all devices and software while scanning and with 650 frames I have done what Revopoint Dual Axis do with 1200 frames , it is just not there yet , but I am sure it going to improve as the results are fine , it just takes double the time to process.

They will need to improve it to make it works with Android phones where the frame count is limited due to RAM , no room to waste any frames .

2 Likes

The turntable connects to the scanner. MINI has built in bluetooth. (discovered removing the bluetooth dongle from the desktop pc, and revoscan would still connect to the table)
Shouldn’t be too hard to “sync” the table position since revoscan knows both the exact moment the scanner sent the start command to the turntable and the speed/angle of rotation.

1 Like

I just made some test and indeed MINI does have now Bluetooth, it is called ZX_POP2 and I don’t need to use Bluetooth on my PC anymore to connect the turntable . It connects via MINI .

The button in the software just trigger the turntable cycles , it do not react to stop of scanning of pause so just one button function .
The speed or tilt meaning nothing , the scan just scanning what it have on the plate , regardless of the tilt or speed .

It would be possible to make the turntable in synchronization with MINI , but that would request the software to evolve including MINI , it would made MINI something else as it is now , most of scanners that are in synchronization with their turntables scanning individually per frame , not 10 frames per second .not even stopping for a second while the turntable change position .
Different kind of animal here.

But hey, I am not against it , if they could make it happen it would be actually amazing and more accurate than ever before , they never stopped evolving so who knows what they going to do next .

2 Likes

What I suggest does not require feedback from the turntable to the scanning software (although that would make the scheme more robust). As long as the software is aware of the rotating speed and angle settings that have been sent to the turntable, it can trivially predict its position delta between any successive frames. This need not be accurate to within a fraction of a mm.

What the software can compute out of this is the position of the camera w.r. the scanned object, for every frame. This is a simple computation, orders of magnitude faster than what is already being done, that can eliminate egregious errors the MINI is prone to, such as suddenly generating a point cloud fragment shifted far away from previous points. And all it requires is a software upgrade; the technical capability is there already.

1 Like

Your suggestion have been forwarded to our R&D team. Thanks!

2 Likes

That would be the way to go , however MINI was designed to be portable hand held scanner that produces as much frames as possible for that process what is really not much but it works . Seamless portable scanners produces at least 80 frames per second to ensure proper tracking but the MINI hardware is too small to process it at that capacity.

Having MINI scanning ability per frame would require a lot of alternations to the firmware ,software and turntable drifting in another direction , you will get another Eindscan in the process what I really did not enjoyed.
On top of the ability to scan per frame , MINI would need to have fixed position to ensure accuracy from frame to frame at the same perfect distance , eliminating any shakes in solid environment and huge limitations regarding the object size .

I think a great idea for a new precise product with 0.01mm accuracy :wink:

3 Likes

@PUTV: You are talking about a whole different product. My suggestion does not imply relying on the turntable to provide 0.1mm accuracy. Approximate camera position would still be useful, as it would serve as a sanity check to prevent gross dislocations of the point cloud. I believe that turntable position can be predicted to about a degree. If you look at what happens to the point cloud between successive revolutions of the turntable, the errors are much larger than that. It was from one of your posts that I learned to never even attempt continuing scanning over more than a single revolution :wink: Relying on the calculated turntable position as a supplementary check would eliminate these misalignments.

A closely similar method is used in photogrammetry, where camera position is determined approximately to prealign successive photos. Yes, I know photogrammetry sucks—I had the bitter experience of experimenting with it. But it would suck much more without relying on the camera position crutch.

2 Likes

@Ralf I just give some ideas if we go in that direction .
I am all good with your proposal here , but major changes need to be adjusted by the core first .

The problem right now is that the algorithms changed drastically, the point cloud results are not acceptable for me , with or without the turntable , on android or windows.
The point cloud on Android was perfection , now it looks messy producing grainy results as it is on windows .

I am honest as much as I love Revopoint products this is not good what is happening at this moment , we going backwards .

So before any new ideas regarding turntables I would prefer that the point cloud is fixed in all apps as it is not usable outside Revo Scan , not even in Cloud Compare. Too many lose points that are not even fused correctly and out of order.

How you going to perform a scanning with Dual Axis having that , it will just create more mess that only meshing in Revo Scan can fix it .

And when people ask why I am converting meshes from Revo Scan to point cloud to merge them , that is the reason , as only Revo Scan can get rid of the mess .

Revopoint Dual Axis was a good idea but the way it all works , it is poorly executed.

The first design of the Dual Axis was perfection , with perfect pivot , then it was changed to less accurate pivot design like the person doing it had no idea about 3D at all , it should be controlled by an separate app, now we got semi automatic turntable collecting additional frames for no reason resulting in even more messy point clouds , wasted double time on processing for no reason.
Where is the improvement?

I did not even made any video yet about that since I feel like telling the truth will not make any good . So I wait until everything have arms and legs again .

I hope you understand now better my point of view on this case .

3 Likes

@PUTV Thank you, now I understand the context better—as well as why I am getting such a lame performance.

Hopefully software upgrades will fix it eventually, although what you say about the pivot is a hardware problem, and here to stay. Frankly, I haven’t noticed a problem with the pivot, but that’s probably because software bugs are overshadowing it. Another hardware problem I found is that there is no way to align the tilt direction with the scanner short of eyeballing it, but that I can work out on my own.

The suggestion about using the predicted turntable position to improve spatial awareness for the scanning software is still valid though, and should produce an improvement independently of other fixes.

2 Likes

The pivot is at the edge , is not as bad but could be better, mainly perfect pivot helps scanning the model more accurately due to the base being in steady position all the time independent from the tilt , let say the origin of the model while it was started scanned and registered in space stays always the same . The first design was actually doing it , and drifting from this idea was not good move , it is impossible to buy a perfect pivot dual axis on the market as everyone just offers exactly more or less what we got now .

And no, we can’t change the hardware , but there is huge room for improvements in the software that can and should be addressed for the best scanning experience.

Everything can be done , for that reason we share our suggestions .

Hopefully sooner than later , until then I am continuing using my YT-1000 dual axis turntable , as it perform the way I need it to .

3 Likes

Reading the logs during a scan (to make an indipended app for the turntable) i saw that once the turtable start rotating, the scanner sends angle requests to the table (every seconds or so).
The table sends back it’s angular position.
So, if the scanner and revoscan both know the position of the table/object why can’t this be used to help tracking ?

+DATA=-12.1;gatt read:+DATA=-12.1;
Notification status changed
Char4 ccc: 0x0 0x0 
D(311021)[main]: 311,SL1:0,SL2:0,KP:0 hxa:-18 fya:30
Notification status changed
Char4 ccc: 0x1 0x0 
D(311443)[GATT CTRLL]: gatt write data:2B 51 54 2C 43 48 41 4E 47 45 41 4E 47 4C 45 3B 
+DATA=-22.7;gatt read:+DATA=-22.7;
Notification status changed
Char4 ccc: 0x0 0x0 
D(312021)[main]: 312,SL1:0,SL2:0,KP:0 hxa:-28 fya:30
Notification status changed
Char4 ccc: 0x1 0x0 
D(312562)[GATT CTRLL]: gatt write data:2B 51 54 2C 43 48 41 4E 47 45 41 4E 47 4C 45 3B 
+DATA=-33.2;gatt read:+DATA=-33.2;
Notification status changed
Char4 ccc: 0x0 0x0 
D(313021)[main]: 313,SL1:0,SL2:0,KP:0 hxa:-37 fya:30
Notification status changed
Char4 ccc: 0x1 0x0 
D(313927)[GATT CTRLL]: gatt write data:2B 51 54 2C 43 48 41 4E 47 45 41 4E 47 4C 45 3B 
+DATA=-46.1;gatt read:+DATA=-46.1;
D(314021)[main]: 314,SL1:0,SL2:0,KP:0 hxa:-46 fya:30
Notification status changed
Char4 ccc: 0x0 0x0 
Notification status changed
Char4 ccc: 0x1 0x0 
D(315021)[main]: 315,SL1:0,SL2:0,KP:0 hxa:-56 fya:30
D(315050)[GATT CTRLL]: gatt write data:2B 51 54 2C 43 48 41 4E 47 45 41 4E 47 4C 45 3B 
+DATA=-56.7;gatt read:+DATA=-56.7;
gatt read:+DATA=-56.7;
D(316031)[main]: 316,SL1:0,SL2:0,KP:0 hxa:-65 fya:30
Notification status changed
Char4 ccc: 0x1 0x0 
D(316170)[GATT CTRLL]: gatt write data:2B 51 54 2C 43 48 41 4E 47 45 41 4E 47 4C 45 3B 
+DATA=-67.3;gatt read:+DATA=-67.3;
Notification status changed
Char4 ccc: 0x0 0x0 
D(317031)[main]: 317,SL1:0,SL2:0,KP:0 hxa:-75 fya:30
type or paste code here

Regarding the Standalone app i’m just using a BLE serial app since last week and i had no problems with it.
That’s the link to a post explaining my process:

I was going to just put an ESP and some mods in the table, but since i saw the scanner requesting to know the angular position of the table i decided to wait hoping for a use of it as tracking helper.

2 Likes

Feedback loop between the scanner and the table would be fantastic for tracking?

But can Revopoint do it?

To tracking what exactly? The Scanner capture the main position in space by its own once you start first frame and follows the object features( by tracking the object featured in real time) to the end of scanning session .

Tracking of the turntable ( object) position would be only helpful if the scanner was frame by frame based scanner , like a photogrammetry scanning style and processing. The features of the object would be no more important and you would be able to scan any free form you want regardless of it’s shape, the way Einscan is working for example .

To do so it would need to be always in the same position of origin what is actually not possible since it is portable scanner and the hardware of the scanner works differently.

For that reason tracking of the turntable position would not add anything to the currently process of scanning or improve it in any way .

1 Like

The position of the table is not calculated by in the scanning process , it all happening inside the hardware of the scanner itself , it do not care in what position the object or turntable is as the tracking happening based of the object origin in the space from the starting point and after following only the object surface and features not it’s position in space anymore .

It is visual based scanner technology and if the turntable tilt or rotate it do not influence anything in the hardware .

I know where you try to go with it , but this is a different scanning technology.

1 Like

I was writing a different answer (i’ll just include it as mumbling) but yeah, you are right.
Everything is done “optically”.
I was actually thinking about the internal gyro assist and how it could have been implemented to assist the scan with weird math functions, but now i’m pretty sure is just implemented as replacing the sensor optical stabilization as in phones.
All is done optically.

You are right… :smiley:

mumbling:

I’m not talking about relying fully on the table position like some scanners do. I was thinking more like an “helper”.
Es. I’m scanning a can.
If i scan it without any markers, it’ll just lose tracking of it cause there isn’t any feature.
Like “we” (revopoint) can track the object spinning from the markers we put on it or on the table, why can’t we do the same knowing the table angular position ?
Instead of optically tracking the table rotation “we” could track the table rotation knowing it’s angular position, since the scanner is asking for the table position anyways. I’m not talking about replacing the optical tracking, but somehow assist it (even just to help realign during fusing) during the turntable mode.

That is exactly that , the gyro helps establishing the first reading of the object in space and that is , it stick to the first position and keep tracking based on the point of origin .
Later when you continue scanning it uses optical recognition, and track based on that remembering the shape and features.

I would love the other way around as the results would be amazing , probably to 10 microns .

The individual frames from MINI are not as good , lots of noises , it needs the motion with multiple frames to calculate the result proper.
The ( not official) DLP laser projector is not good for a single frame scanning as the resolution is not that great either .
It was designed to provide 0.1mm resolution in first place , but I spent long nights to show the team while testing that it can handle much more than that .

1 Like

And thank you for spending those nights with revopoint team, it was well worth it, for all the rest of us😁

1 Like