I’ve calibrated my new Revopoint MINI 2 and its single frame captures are fantastic
But as soon as I start a continuous scan with the dual-axis turntable, I see smearing because the turntable continues rotating while the new image is being captured. It’s only a few %, but for my small CAD components, it is enough to make the scan unusable. This happens both with feature and with marker tracking.
Luckily, RevoScan 5 also has the single frame capture mode. So what I do now is this:
I rotate the turntable about 5 degrees
I capture 2 scans in single frame mode
repeat 72 times
That way, I get much better quality scanning results than with the continuous scan mode and continuous turntable rotation
The problem is, that doing my approach manually is very annoying. And for best results, I would also want to adjust the turntable turning sideways throughout the scanning process. So what I would like to see is an automation function which will do this:
start the turntable
wait X ms
stop the turntable
wait X ms
capture X frames in single frame mode
repeat the above for X times
I’m aware that I can simulate this with an app like https://www.macrorecorder.com/ but I strongly feel like this should be included in the RevoScan software, because competing manufacturers do include such a scanning mode in their software. And it appears to be necessary to get the maximum quality out of my MINI 2.
BTW, in case anyone is wondering if that hassle is truly worth it, here’s an example of the quality that I can get out of my MINI 2 with 1072 single frame captures, which took about 1 hour by hand. The lines in that LEGO engraving are about the size of a hair, yet they are still readable.
BTW I just noticed that the smearing is also visible in your new marketing video. Those marker dots should be perfectly round, but they are a bit longer in the direction in which the turntable is rotating.
This is the reason for perspective.
Perspective is a way of capturing the relationship between depth and proximity of an object in three dimensions, whereas when rendered on a two-dimensional plane, the portion of the object that is away from the observer is compressed, causing the circle to change into an oval.
Hi @fxtentacle
You overdoing it , you don’t need over 1500 single frames to capture that , you already could use continuous mode .
You should capture it all in 90 frames or less .
Overlapping over and over the same region don’t do well for the scan , so less is more .
You did nice job , but don’t over kill it next time , waste of time and 90% of the frames will be eliminated in fusing process anyway because not needed.
Make sure you capture at least 3-4 angles at max .
With so many frames you did on that Lego , I would finish scanning full body person .
My impression so far is that if I take multiple captures of the same scene, that reduces the affect of sensor noise in the Z direction, thereby, making the final scan more smooth.
But you’re right, 1000 frames was probably too much.
Here’s a continuous scan of a continuously running turntable that I just did. The two large things are clay, but as you can see, all the marker dots are elliptical and not round. In this case, that’s clearly not due to perspective, or else the overall shape of the turntable surface wouldn’t be round anymore. Instead, those marker circles turning into ellipses is due to tracking errors accumulating over time. And that’s what doing static captures while the turntable is paused can avoid.
For photogrammetry maybe bit not in 3D scanning , overlapped frames only produces more overlapped and isolated points that induce artificial noises when not cleaned proper , meaning removing more than 90% of the frames you take .
A simple cube needs 6 frames , not 100 , but if you have undercuts in your volume you need to get them from all angles , later the fusing process removing the additional frames that was overlapped , sometimes it leaves some for cleaning after if the object was complex .
But that is how the technology works here .
Overscanning is a huge no !
The markers are not a volume , you need to test it with a proper volume and not 2D flat markers .
I would not recommend using the turntable for single shot mode unless it use the android app where you can set up the timer
You can download it here if you own the dual axis turntable, it works great .
And if future please use proper volume according to your scanner specifications to test out stuff .
That looks amazing for the turntable part, but from what I understand, it cannot trigger the camera capture (yet). There seems to be a GH issue discussing if they can trigger over USB, but my impression is that I need RevoScan to run to collect the resulting frames and then RevoScan will also take ownership of the USB serial port, meaning it cannot be used by other apps anymore.
Wow, that one uses a pretty cool hack to work around the RevoScan software not supporting this: You control the ESP32 via WiFi and towards your PC it pretends to be a Bluetooth keyboard so that it can send a Space keypress to the host PC which will then trigger the single-frame capture inside RevoScan.
Not sure if POP3 is different in this regard. I have tried to scan a piece of irregularly shaped object ( several dried lotus seeds glued together ) with a lot of fine details on the surface. I did a continuous scan ( at “excellent” distance range ) on the turn table using featured mode. The rotation speed is one turn per 30 seconds.
Then I did another scan using single-shot on the rotary table of my milling machine. 36 single-shot frames were taken over 360 degrees. I could not see any “smearing” effect at all. The marker dots are round in both cases and the level of details on the object is the same. It will be interesting if you can post a side-by-side comparison the of the results obtained by the two scanning methods.
The markers are not a volume to begin with so the scanning results of them does not matters.
@new23d you have a very good tracking volume in the middle , that why yours are looking fine .
I do not use the marker plate if I don’t use Marker mode , it only messing up the accuracy in the processing . Marker plate should be only used with Marker mode .
I know marker plates should not be used in feature tracking mode. This is done deliberately just to check whether the marker dots will appear as ellipses in the model due to the motion of the turn table as claimed by the OP.
