Calibration/Accuracy possible with Photoscan?

[3D_Scan_Fan]

Hey guys! Im new here, how are you all doing?

Im a great fan of 3D scanning and in particularly this piece of software on what it is able to achieve which i think it's pretty amazing especially compared to hardware scanning that cost ten times more than what we have here!

In regards to that im really wanting to do somesort of scanning of some parts through Photoscan but accuracy is priority in this scenario, which is probably why hardware scanners might have that advantage for those who pixel peep. I am after capturing objects with something like 0.5% tolerance or 0.01mm+/- accuracy to really get those fine details.

My question then is, is it possible at all to achieve that kind of accuracy with Photoscan by any chance? Is there any sort of calibration that would need to be done perhaps?

I hope you lot understand what im saying or what i am after. I hope all is well and i look forward hearing back too!

[Marcel]

What size are the objects you want to scan?

If you can photograph it with enough resolution, then Photoscan can build it. But it all depends on the size of  the object.

[3D_Scan_Fan]

It's around the size of an Xbox controller, and smaller like the buttons and parts inside it.

Would it suffice?

[bigben]

Calibrate lens parameters
Scaling model is easy enough using control point pairs as scale bars... accuracy will depend on how accurately you measure

The complexity of the shape you are scanning may cause problems for both methods. Scanners have an offset between the laser and the sensor (as do structured light scanners) and as a result shadows are cast into small crevices that result in holes in the scan.  We have a NextEngine scanner (because it was a cheap way to explore laser scanning) but I'd rather use Photoscan in most cases. A good scanner would be nice but that's another magnitude of cost.

[3D_Scan_Fan]

Well i want the accuracy to be strictly 1:1, literally if I was to scan something and then have it printed it has to be 99-100% identical, how would I go about achieving this with Photoscan?

I understand what you're saying about the complexity of the shape, my main worry is the reverse side where there are many inserts and fittings that could be difficult to capture as you can see here in the attached photo, but i am hoping for the best personally. My idea is to capture each photo with long exposure shots to really get the details within the shadows for it to register during scanning.

[3D_Scan_Fan]

Also i don't understand how a scale bar with two points would help me get very precise measurements, Im looking into micrometer accuracy and i don't think your standard dslr would be able to resolve that detail (i could be wrong).

Does anyone know if it is possible to achieve micrometer precision accuracy with Photoscan?

[chrisd]

Does anyone know if it is possible to achieve micrometer precision accuracy with Photoscan?

One thing to note is there is a difference between accuracy and resolution.

It likely that you do not need such a small value for either one for your application. You should determine the minimum acceptable values for your application. This has an associated cost in time and money and requiring a lower value will increase both.
 

[3D_Scan_Fan]

Does anyone know if it is possible to achieve micrometer precision accuracy with Photoscan?

One thing to note is there is a difference between accuracy and resolution.

This is true, however the reason I brought up the two factor was that I was thinking of somehow using a micrometer ruler or chart or anything of the like and place it in my shot when I photograph the object, therefore if the resolution was good enough or even if the lens was capable to resolve micrometer distances then I could potentially use those points on the scale/ruler as reference to set up my measurements.

It likely that you do not need such a small value for either one for your application. You should determine the minimum acceptable values for your application. This has an associated cost in time and money and requiring a lower value will increase both.

I can pretty much assure you I do need such a value, as I have previously mentioned I want tot ry and use this method to make a pretty damn accurate representation of my object that will act as a replacement part if it ever gets damaged, I am prepared to do some work on it but I do need to get it down to that level of accuracy.

Would anyone know if this is possible with Photoscan by any chance?

[stihl]

Isn't an xbox controller copyright protected?

[Marcel]

I think it is possible, but it will certainly not be easy. Smooth black plastic is not ideal for photogrammetry. You would need to add some sort of pattern to the surface to make it easier for Photoscan to reconstruct the shape.

Because it is a fairly small object you probably need to use a macro lens, and you will run into trouble with depth of field being too shallow because you are so close to your object. You can use focus stacking to get around that, but that adds a lot of work (and more things that can go wrong).

Another problem is the deep crevices, Photoscan will have a difficult time to get a good reconstruction of those areas.

The only reverse engineering work I have seen using Photoscan was large scale (large metal parts for ships). The only way to be sure would be to try yourself? It's not just a question if Photoscan can do it, it's also the question if you can create photos that are high enough quality.

[3D_Scan_Fan]

Well i could always use a white coloured controller shell, would that be better perhaps?

I do have a couple of macro lenses that would help for the very small objects, and in terms of the depth of field I could always close-down the aperture and do a capture at a long shutter speed in order to get it exposed. Would this work?

I thought too that the deep crevices might be a problem, but surely if you capture it at many different angles (via. turnable at small increments?) then surely that would provide it enough parallax for it to calculate the difference of depth within them.

And this is not only for this object that I would want scanning, there will be others that will need such precision accuracy, once again how would I go abut achieving micrometer accuracy? How would I calibrate or set it up? for Photoscan? 

[chrisd]

Well i could always use a white coloured controller shell, would that be better perhaps?
...

And this is not only for this object that I would want scanning, there will be others that will need such precision accuracy, once again how would I go abut achieving micrometer accuracy? How would I calibrate or set it up? for Photoscan?

A White shell is much better if you are concerned about the accuracy. Paint or a developer spray might add 0.003" or so, and the crevices in your part might force you to apply more coats to get better coverage, resulting in greater thickness in some areas.

Maybe I have misunderstood what you mean by micron accuracy,     1 micron = 0.00003937 inches (0.001mm)

So if you truly need 1 micron accuracy, what machine are using to create this part?

[bigben]

... and what 3D printer is going to reproduce that accuracy   Easiest way to explore the limits is to start with your calibration target. That will give you a quick idea of what magnification you're going to need and whether or not it's going to be practical.

[3D_Scan_Fan]

A White shell is much better if you are concerned about the accuracy. Paint or a developer spray might add 0.003" or so, and the crevices in your part might force you to apply more coats to get better coverage, resulting in greater thickness in some areas.

Maybe I have misunderstood what you mean by micron accuracy,     1 micron = 0.00003937 inches (0.001mm)

So if you truly need 1 micron accuracy, what machine are using to create this part?

Well then that means I can just go ahead and get a white shell, thus I wouldnt need to do anything like paint of coat it with developer spray then, I am right?

... and what 3D printer is going to reproduce that accuracy   Easiest way to explore the limits is to start with your calibration target. That will give you a quick idea of what magnification you're going to need and whether or not it's going to be practical.

Is this calibration target a way for getting accurate measures? or to do with the camera's lens due to distortion? I've printed these coded markers from PhotoScan (they're like these incomplete circle bars with a white dot in the middle), that's the only tutorial I could find about getting and setting up scale bars and measurements.

Have you guys heard of the ProJet x60 Series Professional 3D Printers??? 

[chrisd]

Well then that means I can just go ahead and get a white shell, thus I wouldnt need to do anything like paint of coat it with developer spray then, I am right?

If its glossy or shiny smooth, that could be an issue for any scanning process. I have not tried using Photoscan for plastic parts (yet).