Scanner Data Samples

[patmo141]

file sent from Italian Stripe-Light scanner.... interested to see your response.

I'll pop it in here later tonight when I get done with all my "obligations" ha ha.

 

[mrcadcam]

file sent from Italian Stripe-Light scanner.... interested to see your response.

I guess many points are closer to 70um?

I'm curious to overall density?

Scott

That's a **** ton triangles, what was the scan time. Laser scanners are good scanners its just the scan times are the factor. Dental scanners are built for speed cause that tooth has to be done yesterday. Point clouds could be enormously larger if the user could get access to actual scanner settings. You can tweak some of the 3shape scanner defs for motor control, but that should be done at your own risk.

What scanner is that if you don't mind me asking?

 

[DMC]

It is for Implants and bars.

Not high productivity. It is for high accuracy and big cases.

It can scan any articulator, big or small.

Maybe it is good for single units? That's not what I am going to use it for.

This system can scan and design stuff that no other system can. Say's me.

Mine will be here in the states this month. First one to US I think?

It's the latest greatest thing for 2012 IMO, as far as scanners and software.

Not the cheapest, not the fastest. That's not what I wanted.

Also, this is a single axis scanner! Funny huh? My choice for the money goes to this. It works great according to my research and sources....we'll see?

Nobil-Metal

http://www.nobilmetal.it/public/products_attach/stampa_sinergiascan_eng_15_02_12.pdf

Some people went to Chicago...others stayed home and did online research and networking on different continents. There is a whole big world out there!

 

[DMC]

Many points closer than even 10um.

I just again picked a random pair of points to measure.


Well...Check it! A pic is worth a Thousand words, or over 6 Million data points!

 

[patmo141]

Italy Scan (Prep Molar Die Model)
vert spacing: 67.1 microns
are density: 287.9 verst/mm^2

Italy Scan (Prep Premolar Die Model)
vert spacing: 65.1 microns
are density: 308.7 verst/mm^2

Italy Scan (Adjacent Teeth)
vert spacing:175.8 microns
are density: 38.2 verst/mm^2

I've also updated the blend file in the original post.

 

[DMC]

Thanks Patrick!

The top of preps are 2x denser, no?

I am buying this for accuracy of scanning implant locator thingys.

Not scanning margins.

I am after a different goal.

Understand?

TOP SURFACES...for alignment of library of implant stuff with the mesh of locator I scanned.. Very very important for bars and implant bridges!

Re-calculate. I dis-agree with your findings! LOL

 

[patmo141]

Fo Sho...



The results varied slightly based on how I selected the top surface so I will include a range
Italy Scan Top Surface
average vertex spacing: 47-49 microns
area density: 605 - 570 vertics/mm^2

 

[Scanner Man]

Just thought I would add my comments here. I work for one of the scanner manufacturers in the market. The vertex density is a poor method to determine accuracy. Within the STL process one can easily re-mesh a stl model and subdivide the vertexes by any factor. We can easily manipulate the model to produce any number of vertices/mm^2 we desire to output.

You can tell if a company has a more advanced stl process if the STL looks similar to 3Shape. The 3Shape stl demonstrates an efficient use of curvature based stl processing. Higher density of triangles in high curvature areas and lower density of triangles in low curvature areas.
We also use this method and the resulting STL is smaller, highly accurate and easier for the CAD to handle.

The Italian Stripe-Light scanner looks to be less accurate even though the system has a high density of triangles. The density on the margin is lower than that of flat surfaces. What this says is the scanner has poor resolution in high curvature areas.

To really check the resolution of a scanner I would make dies with different spacing of grooved patterns. (Like scratches) You can vary the depth and lateral spacing to get an idea of the resolution. Obviously the same die would have to be used on multiple scanners.

 

[DMC]

Metal coin as test piece.

We all know what it should look like. No need to measure.

The density at margin is one of the best yet on scanner sample above.
I took a screen shot of the poorest area, and I think Pat also sampled the same area. Even still, it is among the best.

And the top surface is even better. You say poor huh?
There is no sub-dividing of any triangles on my end.

All scanners can scan a flat surface better than a curved surface that is tilted at angle and at edge of fringe. So,what's new? That's how my human eyes work as well. Same with any camera. No? You have magic-camera technology that allows you to see around corners?

So what if there is even more data on flat surfaces?
Back to the margin, or curved surfaces.....I think this example is just as good or better than anything else in this thread, or anywhere else in the Dental field. To say it is poor is really talking 5hit. I hope you have something to back it up!

I undestand what you are saying. 3Shape was/kinda still is 32 bit and could never do multi full anatomy with all the raw scan data. RAM/Memory overload. Same problem with DentalWings and the Lava platform.

So, mesh was reduced in selected places, and it helped the speed of processing the data, and allowed you to do a little more. We got that part.
If you have 64 bit and decent computer, it is not much of a problem in the year 2012 to have too much data.

 

[DMC]

this scanner collects data this dense in a 200mm x 200mm x 200mm workspace. (Almost 8" cubed)

It is enough to cover any Human mouth on the planet.

Partials, Dentures, Implants, Genitalia scanning maybe?
We had a talk already about not trying to sit on the scanning table, like a sit-n-spin.

You have/know of something better, with this large window AND better accuracy/density?

This scanner hooks up to computer using not One, but Three USB 2.0 cables and a VGA cable.

 

[mrcadcam]

Just thought I would add my comments here. I work for one of the scanner manufacturers in the market. The vertex density is a poor method to determine accuracy. Within the STL process one can easily re-mesh a stl model and subdivide the vertexes by any factor. We can easily manipulate the model to produce any number of vertices/mm^2 we desire to output.

Agreed, most standard cad applications can boost this effect.

*** is this a file you had scanned in front of you, or was it sent to you?

You can tell if a company has a more advanced stl process if the STL looks similar to 3Shape. The 3Shape stl demonstrates an efficient use of curvature based stl processing. Higher density of triangles in high curvature areas and lower density of triangles in low curvature areas.
We also use this method and the resulting STL is smaller, highly accurate and easier for the CAD to handle.

Why not let more advanced users have access to actual scanner settings though instead of having to play with XML and scan def files? Some dont want just speed at every application.

The Italian Stripe-Light scanner looks to be less accurate even though the system has a high density of triangles. The density on the margin is lower than that of flat surfaces. What this says is the scanner has poor resolution in high curvature areas.

To really check the resolution of a scanner I would make dies with different spacing of grooved patterns. (Like scratches) You can vary the depth and lateral spacing to get an idea of the resolution. Obviously the same die would have to be used on multiple scanners.

Penny would be a good place to start. But it needs to be brand new to get the corners of the round surfaces

 

[Scanner Man]

Metal coin as test piece.

We all know what it should look like. No need to measure.

This is not necessarily true. When making the scanner we have to consider scanning time, processing time and dental objects. Most fast scanners in the market fine tune the scan strategy to the object being scanned. For instance Prep's are scanned with different settings than adjacent teeth and so forth. Scanners can get good scans of coins but most do not have settings targeted to scanning a coin.

The density at margin is one of the best yet on scanner sample above.
I took a screen shot of the poorest area, and I think Pat also sampled the same area. Even still, it is among the best.

I was going by the screen capture in the post. If you say that it is better in other areas great but why was it so poor in this area?

There is no sub-dividing of any triangles on my end.

On your end maybe not but but it real easy to do before the stl is saved.
Sub dividing of triangles is commonly used because it allows for better processing in high curvature areas.

I undestand what you are saying. 3Shape was/kinda still is 32 bit and could never do multi full anatomy with all the raw scan data. RAM/Memory overload. Same problem with DentalWings and the Lava platform.

So, mesh was reduced in selected places, and it helped the speed of processing the data, and allowed you to do a little more. We got that part.
If you have 64 bit and decent computer, it is not much of a problem in the year 2012 to have too much data.

In fact this is still a issue in today's systems. To process a 3D point cloud from points to an stl it requires significant processing time. Most processing functions can not use multiple processors because you can not subdivide the cloud and process independently. The 3D cloud requires a single thread for processing of data. Functions such as smoothing, noise reduction and edge sharpening do use multiple processors but the majority of the stl creation process can not. FYI: Our Points to STL engine is 64Bit and will use multiple processors when possible.

Please note that with some scanning technologies the processing of point data to stl can be done by multiple processors if the points gathered are in a structured evenly spaced point cloud.

 

[Scanner Man]

Why not let more advanced users have access to actual scanner settings though instead of having to play with XML and scan def files? Some dont want just speed at every application.

We don't because of frustration on the user side. It is not necessarily true that more data means better results.

Before we went to market many years ago we had this concept. In beta testing the user could choose his own resolution.

The questions from the users were: Why would I use this resolution over this one? When I use the highest resolution why does it take 5 times longer to scan? When I scan a 5 unit bridge at highest resolution my CAD is too slow to use and I wasted all that time scanning at high resolution. I scan everything at low resolution. My copings fit but are slightly loose.

It is not necessarily true that more data means better results. If a scanner is only capable of 20um resolution in X and Y you will get higher noise in the scan if you scanned at a resolution of 10um's in X and Y.

Please note that most scanners in the market have a better Z resolution than their X/Y resolution. Also note that the X/Y resolution is not only determined by the mechanics it is also determined by the measurement technology. If the system uses a laser line the line width and camera resolution needs to be considered. If the system uses structured light the resolution of the camera and projector need to be considered.

 

[patmo141]

The vertex density is a poor method to determine accuracy.

To really check the resolution of a scanner I would make dies with different spacing of grooved patterns. (Like scratches) You can vary the depth and lateral spacing to get an idea of the resolution. Obviously the same die would have to be used on multiple scanners.

Indeed...this has been discussed. But just to reinforce.

If I measure the width of a human hair at 100,000 places along it's length using my eye and a millimeter ruler would you say that I knew it's shape more accurately than if I measured it in 10 places using laser diffraction or a forensics microscope?

How many places I measure is resolution
Accuracy is determined by the tools that I use.

Now, to put this in the context of a stripe light scanner....

The number of pixels in the image sensor will determine resolution

But the accuracy is going to depend on
-clear, distortion free optics on the camera and projector
-signal quality in your image sensor
-extremely good calibration
-stable environment etc etc.


To represent small features, like margin edges, grooves, implant interfaces etc....

You need accuracy and resolution. And one does not imply the other. Scanner man raised a good point, does taking measurements at smaller spacing than your uncertainty make any sense?

 

[DMC]

Uncertainty?

How do you measure that?

What are you talking about?

 

[lcmlabforum]

I vaguely remember the basic science class with our well esteemed profession
of biomaterials in Houston and Ann Arbor (having PTSD from mentioning people
by name on the forum).

There is a limit to the accuracy if your measurements is only up to 0.1 mm.
So if your tool can read either 0.1 and 0.2, 0.3, saying a distance is 0.05
means it is a guess and there is the concept of what is the error value.
So if your measuring tool, like a laser, can only measure up to a certain
level of precision, repeating it 10 x is not going to make it more accurate.
Precision is htting the target the same place all the time, but may be off
10% all the time. Accuracy is hitting the target nearer the center, but
the distribution is possibly more dispersed. We have a term in target
practice - I think it is call the spread . . .
Patmo - were you referring to that?
LCM

 

[mrcadcam]

We don't because of frustration on the user side. It is not necessarily true that more data means better results.

Yeah but when the first scan is a turd, numerous times when scanning a full arch(Lingual side of the teeth there are voids or mis-scans). Most of the data is usable from that 1st scan except for that area. Instead of having to rescan again, merging 2 point clouds is more managable or letting us in real time correct scanning attributes for that particular scan would be plus.

I realize 3shape has to dumb it down for masses, but were not all dolts.

Before we went to market many years ago we had this concept. In beta testing the user could choose his own resolution.

Brilliant,

The questions from the users were: Why would I use this resolution over this one? When I use the highest resolution why does it take 5 times longer to scan? When I scan a 5 unit bridge at highest resolution my CAD is too slow to use and I wasted all that time scanning at high resolution. I scan everything at low resolution. My copings fit but are slightly loose.

3shape was obviously beta testing with people that have no computer knowledge or common sense. That sucks the software is being built around these users.

 

[Scanner Man]

FYI - I am not with 3Shape

In the market you can purchase scanners not intended for Dental that will allow you to do what ever you want.
But you will quickly realize that dental scanners have an optimized workflow to make the scanning more efficient and provide the accuracy required.

As the Dental Lab scanner market matures I think that you will see more advanced systems that have this user flexibility but the average user today still does not have the expertise to make all the decision. I would suggest to always provide feedback. I know that if enough users request something it will eventually make it into the software.

Going back to accuracy. The different technologies and system designs in the market have different strengths and weaknesses. The technologies and designs for scanning dies is not necessarily the best for scanning implant locators. The best for implant locators is not the best for scanning dies.

In general the better systems for locators tend to keep the model in one position during the scan and don't have to move many axises. The output should not be a stl but instead be a vector or vectors with an implant seat plane.

The better systems for scanning dies tends to be one that can move the model in all different angles to capture as much of the model or undercuts to generate a accurate 3D model without using hole filling.

 

[NicelyMKV]

Scanner Man, nice to have you.

I was talking about the differences in red laser and white light scanning technology in Chicago. It seems like red laser scanners can be negatively influenced by a greater number of variables than white light.

Object color, texture variations, highly reflective surfaces, ambient lighting and translucency to name a few. White light technology does not suffer from these same issues.

What are your thoughts on that?

 

[patmo141]

I vaguely remember the basic science class with our well esteemed profession
of biomaterials in Houston and Ann Arbor (having PTSD from mentioning people
by name on the forum).

There is a limit to the accuracy if your measurements is only up to 0.1 mm.
So if your tool can read either 0.1 and 0.2, 0.3, saying a distance is 0.05
means it is a guess and there is the concept of what is the error value.
So if your measuring tool, like a laser, can only measure up to a certain
level of precision, repeating it 10 x is not going to make it more accurate.
Precision is htting the target the same place all the time, but may be off
10% all the time. Accuracy is hitting the target nearer the center, but
the distribution is possibly more dispersed. We have a term in target
practice - I think it is call the spread . . .
Patmo - were you referring to that?
LCM

yes, this is exactly what I'm talking about. Sticking with the target practice analogy. Uncertainty could be described as...

"If I aim my scope at the bulls eye, I am confident that the bullet will hit the center of the bulls eye +/- 2inches 75% of the time and +/- 5 inches 90% of the time and +/- 8 inches 99.99% of the time"

For a scanner to claim "20 micron accuracy," it means that for every single data point in the mesh, it could be where it says it is +/- 20 microns. But it's really not as simple as that. As Scanner Man hinted, this uncertainty can be different in different directions. E.g. +/- 20 microns in the x or y, direction but only +/- 2 microns in the z direction.