Scan with Mini, question from newbie

[Harh]

Hello

First thing I was to scan after the Revopoint Mini has came (after successful scanning of the bust that came along in the box), was a Hubsan Zino 2 quadrocopter arm, that one of my friends asked me to model.

As you can see, it is a long thin object, that is already white matte plastic and that has completely refused to scan using turntable, when I’ve set it just lying on the surface. It was due to the fact, that it couldn’t just properly fit in the “too near-too far” range of the scanner.

Upon reading the forum, I’ve got that I should scan it vertically in a few parts and then combine them. I’ve added a few markers, that you can see on the photos up there, and also added some black dods, thinking it wil help tracking the object properly… Only thing I’ve got is some kind of very persistent impression, that Revopoint tracking isn’t that bad at all, but it is an ABOMINATION THAT IS MADE FOR PEOPLE TO SUFFER GREATLY! )) I understand, that PUTV, whom I’ve already learned a lot of respect due to her tremendous work put into the scanning and helping people here, might not agree with me, but still… But what am I… Let me be more constructive…

I’ve got a few scans, but their look wasn’t any kind of promising, and I just couldn’t get them combined since scanning those with the arm set vertically, I just can’t get a scan with enough points of both top and bottom parts to combine them. I’ve split the scanning into 3 parts, but the problem with aligning those persist.

I’ve tried to handscan that arm with it just lying down the surface along with some small cylinders and other stuff like that for the scanner not to loose tracking, but this way I either make “one and a half” sides of the arm, or I need to repeat the scanning with another angle or try to turn the scanner above the object, but both of those give problems with tracking - it doesn’t loose it, but the points get aligned wrongly, that greatly affects the result. Sorry, I’ve forgot to make a photo, but here you can see the general idea of how it was scanned

Thus I’ve got 4 scans that represent one side of the object and one another (side one) to some degree. You can see those scans already looks suspicious with their far left and right parts already starting to blur.

Upon combining those in Cloud Compare, I’ve got a result, that I can immideately print just to be sure I have something physical to throw out of the window ) Alignment was done by rotating and translating the scans cleaned of the parts added fo tracking purposes, so the pair would visually fit, then just pressing that button:

The final mesh looks this way:

There is obvious mislignment problem, though I am unsure did it happen by the combine made in the Cloud Compare or that was already in the scan data (looking upon the point cloud in RevoStudio it more looks that was already there). The general result, you see it by yourself.

May I ask you to point me out, what was done wrong to the degree, that led to this result and how to avoid it? BTW I’ve calibrated the scanner before - forgot the exact numbers, but they were not worse that people here gets upon the calibration.

[JeffLindstrom]

The first thing that comes to mind is that it appears you did not have enough markers in each frame (there have to be at least 6).

I think you would have better results if you added small objects around the object to provide more unique patterns in Feature mode.

[Harh]

I’ve used feature mode. Markers here are mostly to make flat surface be more distinctive. Yes, I’ve tried marker mode, and it has failed due to not enough markers present. Though I am unsure, is there reason to add more markers since I can add no more than one to three sides of four (fourth it round).

As you can see, I’ve added objects to use them as tracking dots. Do you think that adding more will really improve things?

[PUTV]

You see , you could scan in color mode , and export the meshed part so you can use the RGB data to merge the scans more precisely since you painted so many dots , they are great as marking points for merging .

And yes I do disagree
You could use marker mode and the markets on top of your turntable , keep the MINI slightly above the object at 45 degrees. You would be able to capture it in 2 parts .

Or add additional elements to the object and use feature mode , you can later remove the extra objects before merging .

You asking why there is shifting after merging ? You see MINI was made for scanning small objects , if you scan the same object with different distances even slightly, you getting different resolution/accuracy , so pay attention that the object parts you are scanning are always at 10 or 20 cm from the MINI yo the center of your turntable .
So you get the same accuracy and resolution . Do not over scan areas if possible.

So … since you have a 3D printer , you can print out some small cubes , maybe around 1 x 1x 1 cm and place them around the object on turntable, closer but not touching the main object , it will help you to stabilize the tracking and make sure the ( Remove Plane ) option under Depth Camera is Off .
Scan the first part of the object one rotation 360 degrees, then the other part , then flip it and do the same with the other side … remove the cubes before merging using lasso in Revo Scan or CC .
( You can still use the painted mini dots as marker reference for merging , when you scan in color )

BTW I did enjoyed reading your process as it show me that you did everything the way you should do … and pesky tracking issues are on the object …
Now you can throw out the window the printed version to release your stress and then try again . I am sure it will works .

[Harh]

Seems I’ve finally got it scanned, and since I’ve wasted a lot of time in the process, wasting some more won’t really make much difference. Hope nobody gets insulted and me banned for good…

Instead of an epigraph:

So, if you are a newbie, what do you need to make a scan of quadrocopter arm? Revopoing Mini, 10-12 hours of free time to bloody sacrifice it, RevoScan, Cloud Compare, 3dsmax/Blender and a PC to run that all on. I would prefer to have Revopoint tracking programmer at hand, also for some bloody sacrifice - even if it won’t help much, at least it might cheer me up - but alas…

Finally understanding, that software is not really my friend, but in most cases quite contrary, I am were to struggle it till I’ll gonna overcome it - even I’m not cleverer, at least I am more stubborn. So:

1. Adding one pair of markers in addition to those I have on that arm already;
2. Setting the arm on the platform so the complex shaped part of the detail is in the center of it (second group of scans with another end of the arm set in the middle);
3. Getting Mini scanner in it’s place about 16 cm from the detail (diagonally) and tilted roughly 45 degrees down;
4. Setting brightness to 3 and mode to Marker;
5. Speed of the turntable was set the way it make a full revolution in about 40 seconds (360 scanning frames). It was just set that way - not because it is optimal, just it happened so;
6. Also I’ve added some 6 mm 3d-printed boxes. Don’t really think they are needed in the marking mode, but just in case. By the way, sometimes the scanner thought they were markers even they are square, and not round:
   
   
   
   Have a thought that they could affect the tracking in negative way, but only tests can find it out.

The set looks like this:

So pushing the Start button… By the way, I really recommend using theese kind of yellow glasses, when working with the scanner, since it eliminates that freaking blue flashing almost completely:

Even maybe it is not that bad to one’s eyes, it is really annoying. Especially due to the fact you can’t turn depth camera off, when RevoScan is open, and you need a few tweaks/rearrangements before next scan. Would recommend Revopoint to add this function.

This way I’ve made 9 scans: 4 sides for each end and 1 to control the final “stitching” process with tha arm centered by the middle of it. In fact I’ve used only 6 of them - 3 for each side and discarding the “intermediate” one.

Considering current scanning processs, there was nothing interesting, just a lot of time scanning and cleaning the scans after.

By the way, why I keep blaming the software… This is a Marker Mode. So when the markers are static, clear, clean, white, black, obvious and plain. And still I keep getting this after the one table round of scanning process:

Overall image is clear, there is a marker to the left, and to the right down there you can see that marker from the “left view”, I mean when it must be completely planar, but it has about 400 micron thickness, that is a bit excessive I think. Maybe it is due to those cubes, will check that, but for now I doubt they are the case. And with this Revopoint Mini was able to scan that engine part (or whatever it was), really? Maybe it was that AESUB spray, that, considering it’s price, must contain a good ammount of black magic in it, - rumors says it does, when you try to coat a copper coin with it - though I doubt it too.

At first I’ve tried to align them using multiple point alignment in CloudCompare like this:

But the results were bad as you can see in the first post of mine up there, obviousely I didn’t get “make ends meet” out of those scans.

Interestingly, the problem still may be in the fact, that I’ve used black dots, which I hoped will help tracking process in the Feature Mode, but found out they don’t help at all. And using them to aling the model, since they are visible in the point cloud, might be a bad idea. I’ll show why a bit later.

So the alignment was done completely manually in the 3dsmax. Don’t ask, how much time did I waste. Maybe I should rescan it in color mode as PUTV said up there, removing those dots first and adding new ones of different color, but I just moved the hard way since work was already done, and I didn’t want to start over again.

Considering final cleaning and alignment, I can tell you only there was a moment, when I looked upon the detail, it looked upon me, then I say “Have I finally did it?!” with it answering “Freaking yes!” in some kind of bad voice (yup, I stole that joke). And that’s all I remember.

Here you can see the final alignment with different scans in different colors:

One thing to note though. You can see that black dots I’ve made, trying to persuade the software to use it as a tracking, are visible to it not as color things or holes, but as bumps, though it was just a gel pen with overall zero thickness. And you can see to the right of the image, that on different scans they appears at different positions even they are completely static. Maybe that was the reason that multi point alignment didn’t work good, but this is a case to check in future scans.

The final mesh looks this:

Some quiet perfectionalist voice said “Hey, there is some obvious misalignment on one of the sides” following me and the detail itself bellowing “To hell with it!” in the same time.

BTW would recommend to add in the “Isolation” section of the RevoStudio “Keep markers” button, so the central parts of the markers won’t be deleted by using it. Though unsure is it easy to implement or not - markers produce obvious 5 mm diameter flat point cloud patches, but does it really help or not to make this, unsure.

Final thoughts: it seems the turntable produces some vibration due to either gears used or using of stepper motors with not enough substep division (didn’t disassemble to look closely), so it may produce some bad influence on the scanning process (one of the tracking object I’ve added began to vibrate, so I’ve removed it) and there is some turning speed, that is the best one, that may reduce vibration and having good enough exposure at the same time, but I can’t effectively test it. Yeah, PUTV has mentioned it,but I can’t neither find, nor want to test.

Secondly, gonna buy some zinc oxide to try to coat the detail scanned with it using airbrush, hope it won’t get clogged, since the AESUB due ot it’s price and availability is not really an option, but it might or might not be a different story. And thank you, those, who read it to this point and manged not to get asleep. Cheers.

[PUTV]

Wow , that was hard core !
Why you used marker mode with your printed cubes ? If you used marker mode you should stick the markers on top of the cubes ti have better tracking .
Anyway you did it …

Next time scan in color , use blue sharpie for painting dots that you can use for merging as merging points, it always works .

And get rid of the markers on the object , they are not really needed if you use the other one .

To be honest I don’t like to scan using marker mode , I avoid it …I prefer to add some elements and scan in feature mode if the object is small enough .

Here how my marker cubes looks like
Black to use with markers, and white to use with feature mode .

[JeffLindstrom]

@Harh

AESUB spray… when you try to coat a copper coin with it

PUTV mentioned that she discovered corrosion on some of her antique coins some weeks after using AESub to prepare them for scanning.

After using it on several metal objects, I washed them thoroughly to avoid the same result.

[PUTV]

@JeffLindstrom

PUTV mentioned that she discovered corrosion on some of her antique coins some weeks after using AESub to prepare them for scanning.

Yep AESUB Orange in particular, it has finer layer thickness but sadly ruined some of my metal coins , copper, bronze , even a us quarter was oxidated after badly .
The Blue was less corrosive.

I switched now to Attblime and in general using ABR2 and the best ever, 3 microns layer only , cheap in a can when using airbrush …
I will never purchase vanishing spray from AESUB that’s for sure . If you use MINI beside AESUB Yellow ( very expensive) , everything else provide too heavy layers .

For the rest zinc oxide as usual .

[Harh]

Why you used marker mode with your printed cubes ?

Thought it will be more solid tracking solution. I think I’ve tried it with feature tracking, but it trashed it again, but I’m unsure since I’ve got a bit confused with all those numerous scans - need to check it again.

If you used marker mode you should stick the markers on top of the cubes ti have better tracking .

Thanks, will pay attention to it - maybe there is really a reason to make some “eternal black marker dices” with 3D printer

Next time scan in color , use blue sharpie for painting dots that you can use for merging as merging points, it always works .

Yes, that’s what I plan to do next time.

And get rid of the markers on the object , they are not really needed if you use the other one .

As I said, yes, shall try to use feature mode, but still have a real doubt it will succeed, though I should try it anyway. Case I doubt is things I am interested to scan aren’t miniature figures, plants, leafs, oranges and stuff like that - that have a lot of microdetails and so. Smooth plastic surfaces. I don’t think they’ll get a lot better than this. Unsure, maybe color scanning will force Revo Scan to use color camera for feature tracking, but doubt it.

I prefer to add some elements and scan in feature mode if the object is small enough .

For now have no ideas how to make additional “temporal” features to the object fast enough.

Making a render for that quadrocopter arm as a show of the result.

[PUTV]

@Harh

For now have no ideas how to make additional “temporal” features to the object fast enough.

Crumble some painter blue tape in small balls and it will do it , make sure they do not change position while scanning , the cheapest fastest solution , you can use even regular paper as long it don’t move while scanning .

@Harh

maybe color scanning will force Revo Scan to use color camera for feature tracking, but doubt it.

Scanning colors do not track anything , only features in space of an objects .
The projector inside the scanner projects pattern a vertical lines , so the additional objects disturbs the pattern and tracking occurs and get registries in 3D space .

I told you to paint a color dots , so you can use it as markers for merging later the objects , reference markers not tracking .

Color reference is placed exactly on the points of your point cloud or mesh , so it is highly precise when used as reference point for merging .

Just do not create textures … as it will remove the RGB color data from the mesh

Looking forward to see your improvement on that as your time allows , scanning is very labor time consuming, that why 3D scanning is very expensive.

[Harh]

@PUTV

Crumble some painter blue tape in small balls and it will do it

Hmm… Thought that might produce a lot of sticking out thin paper pieces, that might not look good to the scanner. But I’ll try, thanks.

@PUTV

Scanning colors do not track anything , only features in space of an objects .

Yes, I have understood it already now. And that I need blue pen and not a black one.

BTW, made a render of that scanned model:

Quadrocopter Arm