3d scanner

From the update

Quote:

---

To be clear, this update is not about the Scanner Attachment, which was an add on that we also offered. You don’t need the scanner attachment to use the Peachy as a 3D scanner. The Scanner Attachment is simply an add on that includes a mounting system, a camera, and a motor driven scan platform to make for easier and more effective scans.The Peachy Printer kit will include everything you need to use it as a 3D scanner except for a camera. Pretty much any camera that you can connect to your computer will suffice, higher the resolution the better.

---

From what I can see the scanner needs a motor turntable of some sort to turn the object being scanned. And the ppk will include that? I wasn't expecting a turntable. I think I'm misundestanding something.

Quote:

---

Originally Posted by iplayfast

From the update

From what I can see the scanner needs a motor turntable of some sort to turn the object being scanned. And the ppk will include that? I wasn't expecting a turntable. I think I'm misundestanding something.

---

I don't think you are misunderstanding. I think they are mis-communicating. It's not possible to actually scan more than a single line with out rotation.

You do need to rotate the object your scanning.
The 100 dollar printer/scanner dose not ship with a motorized turn table but instead will ship with a simple way for you to move the turn table very smothly by some other means.
We have come up with many many ways to do this but before I post about them Id be curious to know what the inventive minds here at 3d print board can come up with.

Quote:

---

Originally Posted by rylangrayston

You do need to rotate the object your scanning.
The 100 dollar printer/scanner dose not ship with a motorized turn table but instead will ship with a simple way for you to move the turn table very smothly by some other means.
We have come up with many many ways to do this but before I post about them Id be curious to know what the inventive minds here at 3d print board can come up with.

---

http://www.ebay.co.uk/itm/Synchronou...FrWFoSeWKOX4zg

It would be helpful if you guys provided a bit more info like: will the line laser always be stationary or will it have to be moved in order to get the line where its supposed to be? It's my understanding that the laser in its normal printing position is at 90 degress to the platform surface. Is there a mecanism for consistency getting the line laser to make a 45 degree angle as you describe. I think these questions need to be answered before any real solutions can be sought from the community.

it is comparitvely easy to relocate the laser in relation to the camera, and as for having consistancy between the laser source and camera that can be done in the frame/holder/case that the scanning happens in, BUT the big worry for me is the higher power near UV laser scattering off the scanned object and affecting the space around the scan volume, I suspect that this worry is why most of the laser based 3d scanners use low power red lasers

Quote:

---

Originally Posted by rylangrayston

You do need to rotate the object your scanning.
The 100 dollar printer/scanner dose not ship with a motorized turn table but instead will ship with a simple way for you to move the turn table very smothly by some other means.
We have come up with many many ways to do this but before I post about them Id be curious to know what the inventive minds here at 3d print board can come up with.

---

If you had an old style record player, and had a way of tracking how many times a second it rotated, you could then take a picture at the precise interval that the turntable would be rotated at any degree.
Also if you took a series of pictures of something rotating on a turntable, and found the "degree of closeness" to one image you would get a sign wave where the peaks would be where pictures were exactly alike. That would determine the period of the turntable.

Quote:

---

Originally Posted by iplayfast

If you had an old style record player, and had a way of tracking how many times a second it rotated, you could then take a picture at the precise interval that the turntable would be rotated at any degree.
Also if you took a series of pictures of something rotating on a turntable, and found the "degree of closeness" to one image you would get a sign wave where the peaks would be where pictures were exactly alike. That would determine the period of the turntable.

---

Nice... so for the scanner turntable in the campaign video I did something very similar... I was pretty dam broke at the time so I just used the stuff we had around the whorehouse ( I was pretty much living in a where house lol )
I took my brothers skate board and mounted a dumbbell weight to one of the wheels. Next I recorded a video of the paper crane spinning on the platform. It was easy to edit the video so that I had just one rotation, and then turn the video frames into pngs.
If I had 100 pictures then 360/ 100 = 3.6 .... each picture was 3.6 deg rotated from the last. Love you idea to find sine waves from high spots on the model! I had not thought of that, It would be a great way to automatically synchronize multiple scans of the same object too.
If there are multiple high spots out of phase then for some objects there will be unique phase offsets for spots on the model, and further more placing an object off centre on the platform may simply add one big carrier wave that all the other sine waves ride on.

Quote:

---

Originally Posted by harpo99999

it is comparitvely easy to relocate the laser in relation to the camera, and as for having consistancy between the laser source and camera that can be done in the frame/holder/case that the scanning happens in, BUT the big worry for me is the higher power near UV laser scattering off the scanned object and affecting the space around the scan volume, I suspect that this worry is why most of the laser based 3d scanners use low power red lasers

---

Yes not only that but the uv laser light is sometimes filterd out by a plastic film placed over the cameras sensor,
There are many reasons behind the current plan, which is to ship both a Red laser diode and a uv laser diode.

Quote:

---

Originally Posted by tarheelbandb

It would be helpful if you guys provided a bit more info like: will the line laser always be stationary or will it have to be moved in order to get the line where its supposed to be? It's my understanding that the laser in its normal printing position is at 90 degress to the platform surface. Is there a mechanism for consistency getting the line laser to make a 45 degree angle as you describe. I think these questions need to be answered before any real solutions can be sought from the community.

---

The software will likely just let you tell it what angle you have the laser at relative to the camera, and you may need a protractor. This is good because there seems to be a trade of between depth perception and resolution which is controlled by laser angle. At Peachy we like to make our products easy to hack, so many attributes like this will be easy to change later.
You make a good point about needing more info to provide specific solutions, but right now the scanner hardware is not finalized, so this is a good time for you to tell us how you think it should work, instead of me telling you how its going to work.
Of course we are close to shipping so some things are concrete, for example Iv got thousands of Red laser diodes paid for and in the mail, so Now is not a great time to suggest using green laser diodes.
The circuit board were already shipping has the ability to turn on and off multiple laser diodes, and we can update the firmware to the board after shipping.
The ABS plastic has been in my shop for quite some time, but we cut the parts plate just in time( LEAN manufacturing)
So the laser cut scanner parts can still change.

So in short how would you like the scanner to work knowing that there is no motor and there is 1 or more red line lasers that you can turn on and off?

Quote:

---

Originally Posted by rylangrayston

So in short how would you like the scanner to work knowing that there is no motor and there is 1 or more red line lasers that you can turn on and off?

---

I have no idea, but would like to say I approve of your approach. This attitude, and the original idea, is why I joined the funding effort.

If you want to do the rotation cheap and manual it would be quite easy to rig something up since we would have access to a 3D printer, you could just print a gear like structure as the platform and rotate it manually one step at a time maybe use a nail or something to stop it at a precise spot.

Hey guys,

considering you've got the printer setup and can capture around 30fps per sec, I would suggest to use the mirrors of the peachy and make one of them oscillate really fast (or use some other way to get the laser to draw a full line in each videoframe).

Then I would use the second mirror to slowly tilt the line to the left and right to get a distance value for a full 2D grid.
Then you can rotate the object by a step and repeat the process.

This would work similar to many other 3D scanners (eg. Kinect v1) which project a certain pattern on the object and use a camera at a certain distance to get the 3D information.
https://www.youtube.com/watch?v=fJpNHBfLfdU
This 3D animation shows how they use patterns to do the scan. Our "pattern" would be the time difference of the laser-line.



This would have the advantage that, as we will always get a full 3D-picture, the rotation steps can be a lot larger and easier to do manually.

greetings, quertz

Quote:

---

Originally Posted by quertz

Hey guys,

considering you've got the printer setup and can capture around 30fps per sec, I would suggest to use the mirrors of the peachy and make one of them oscillate really fast (or use some other way to get the laser to draw a full line in each videoframe).

Then I would use the second mirror to slowly tilt the line to the left and right to get a distance value for a full 2D grid.
Then you can rotate the object by a step and repeat the process.

This would work similar to many other 3D scanners (eg. Kinect v1) which project a certain pattern on the object and use a camera at a certain distance to get the 3D information.
https://www.youtube.com/watch?v=fJpNHBfLfdU
This 3D animation shows how they use patterns to do the scan. Our "pattern" would be the time difference of the laser-line.


This would have the advantage that, as we will always get a full 3D-picture, the rotation steps can be a lot larger and easier to do manually.

greetings, quertz

---

Quertz..... a narrow mesh (generated by both mirrors oscillating), is an excellent idea. As a matter of passing interest, someone invented a bicycle lamp which projected a mesh onto the ground to identify potholes !!!!! A square mesh would probably suffice.

@mike_biddell

I seriously just read "to identify plotholes" and thought: that's not how this works :D

Actually I rewatched the video of the update and it seems to me that they are planning to do something like that.
I'm just not sure whether they are aware of the similar solutions that already exist.

I would definitely suggest making the angle between laser and camera even smaller. I think the optimum is somewhere around 20 - 30°.
You will get the resolution back by getting the whole 3D image and not just a single line.

Greetings,
quertz

Not sure how much help it will be, but I've heard other 3d scanners highly recommend trying to coat the item to be scanned in a white powder like talcum powder to get uniform reflection and avoid the light absorbtion problem.

Great Ideas everyone!
Iv tried using a very old version of the peachy to move the laser in a line quick enough to simulate a line laser lens.. its a great idea and it still has potential.
one of the problems with doing that is the that typical web cams dont give you control over the exposure time, and the exposure time adjusts automatically, so its tricky to
get expose a line that dosent overlap uneven brightness. The first scans shown in the main campaign video did use a peachy printer with one mirror vibrating very quickly. and as a temporary solution I used a DSLR camera which allowed me to control the exposure times and minimize this problem.
I like the idea of a scanner hack where a line laser ( with a line laser lense) is placed in the peachy printer head and one of the mirrors causes the plane of light coming from the laser to move across the model.

We also wrote software quite some time a go that takes N video frames and mixes them together to create a long exposure shot... for example filming at 30 frames per second, you can take 60 frames to get a 2 second exposure image. This could be useful when drawing a grid.

I Think the first version of the scanner will be simpler than these ideas but I can see how as scanner hack today could become the standard 2.0 scanner of tomorrow.
oh and something to stop it a precise spot... yes Ill see if I can work that in today, seems useful.

Quote:

---

Originally Posted by rylangrayston

Great Ideas everyone!
Iv tried using a very old version of the peachy to move the laser in a line quick enough to simulate a line laser lens.. its a great idea and it still has potential.
one of the problems with doing that is the that typical web cams dont give you control over the exposure time, and the exposure time adjusts automatically, so its tricky to
get expose a line that dosent overlap uneven brightness. The first scans shown in the main campaign video did use a peachy printer with one mirror vibrating very quickly. and as a temporary solution I used a DSLR camera which allowed me to control the exposure times and minimize this problem.
I like the idea of a scanner hack where a line laser ( with a line laser lense) is placed in the peachy printer head and one of the mirrors causes the plane of light coming from the laser to move across the model.

We also wrote software quite some time a go that takes N video frames and mixes them together to create a long exposure shot... for example filming at 30 frames per second, you can take 60 frames to get a 2 second exposure image. This could be useful when drawing a grid.

I Think the first version of the scanner will be simpler than these ideas but I can see how as scanner hack today could become the standard 2.0 scanner of tomorrow.
oh and something to stop it a precise spot... yes Ill see if I can work that in today, seems useful.

---

A bit of Googling..... structured light scanning looks to be the way to go, projecting a series of grids onto the object. Whether Peachy could achieve frame rates for persistence of vision (25-30 fps) I dont know. My cheap laser pointers come with a series of lenses, so a cheap 'grid' lens should be achievable, if Peachy cant achieve a persistence of vision grid.

The David scanner uses a projector to create the grid:- http://wiki.david-3d.com/david3_user...ructured_light

errr this laser pen has a lens which produces an interesting pattern of squares:- http://www.ebay.co.uk/itm/New-Red-La...cAAOSwMmBV5qMF

http://www.ghoststop.com/Laser-Grid-...-greengrid.htm

Since the peachy can turn 4 line lasers on and off, this scanner does that and produces good quality scans..... the main scanner rig to hold the camera and line lasers could be printed by the Peachy itself. Using 4 lasers at different angles to the scanned object seems to give good quality scans. http://www.instructables.com/id/Build-a-30-laser/

I think you could get a really good quality using the peachy for structured light / laser scanning
https://en.wikipedia.org/wiki/Struct...ght_3D_scanner

I've daydreamed about glasses free 3D tv's and wonder if my day dreaming might relate to this somehow.


Since 3d TV with glasses often requires flickering filters in the wearers lenses to show/hide the left or right frame. I thought what if your TV was able to shake left and right and only displayed the left or right frame (or any in between). You might achieve a good sense of depth.


so maybe there is a way to either shake the camera or model on an axis as it rotated to "see" better into the nooks and crannies. and that way you're not just scanning the same section over and over again trying to get higher resolution, you can attack it from more sides. the system would have to know the exact location of everything to account for the shake, thankfully thats not my problem :).

Quote:

---

Originally Posted by ivin

I've daydreamed about glasses free 3D tv's and wonder if my day dreaming might relate to this somehow.


Since 3d TV with glasses often requires flickering filters in the wearers lenses to show/hide the left or right frame. I thought what if your TV was able to shake left and right and only displayed the left or right frame (or any in between). You might achieve a good sense of depth.


so maybe there is a way to either shake the camera or model on an axis as it rotated to "see" better into the nooks and crannies. and that way you're not just scanning the same section over and over again trying to get higher resolution, you can attack it from more sides. the system would have to know the exact location of everything to account for the shake, thankfully thats not my problem :).

---

Ivin

Your dream makes sense, the difference between left eye and right eye view or two separate and distinct view is called parallax, which gives depth perception. This is used as a 3d scanning technique.

Hey guys,

I've actually already tried that.
There is free software out there which you can use to turn a lot of pictures into a 3D-model, but it doesn't work that good.

Peachy would hopefully be great for small objects like those you probably want to replicate using peachy

On an entirely different note, I had an idea while playing with the 3D scanner I already have. Rather than positioning multiple cameras or moving one camera to multiple positions, have a set of large mirrors placed parallel to each other at opposite ends of the scanning area. Orient the laser line to project exactly perpendicular to the mirrors, so that the laser and all of its reflections form one line across the object to be scanned. Then orient the camera to view the line at on a 45 degree angle from the plane of the laser line. Place a visible black line along the bottom edge of the mirrors so that the software can calculate where each line segment ends and have the software overlap and flip the segments as needed to get possibly dozens of similar but different angles of the exact same laser line to fill in gaps that would normally happen from overlapping segments.

Then go completely crazy, have four mirrors oriented in a square box, with two differently colored lasers and the camera at an isometric angle to the box, to get hundreds of differing angles on the same part with each line of each laser.

feign, it sounds good except for one fact, the object to be scanned BLOCKS the laser beam path, so the laser does not hit the mirrors on the other side of the object, but yes offseting the camera in both the x & z directions while still pointing at the centre of the scan space should help scanning the objects, but it would probably need two cameras(one on the plane of the laser(s), and one above the laser plane

I guess I didn't clarify, the lasers should be oriented above the scanning, not even with it...
But then, I'm picturing a very wide fanning laser line. I don't know if a basic lens could really get the kind of spread I'm picturing here.

Hey Feign
Have you built some of this?.... it seems to be a new way to scanning that I havent seen documented anywhere.
Ive read you post quite a few times and I think Iv got it modled right in my head now.... hold up this would take just a few min in blender, im just going to go do it...
Attachment 7860

ok so something like that ... where the yellow box is the peachy printer and so using the mirrors you can hit the object from ALOT of angles.
and obliviously the laser wouldn't keep going threw the object being scanned like it dose in this screen shot.

It seems to me that its possible that you don't need to rotate the object smoothly, instead you could just re position it 3 or more times.

rylan, that is an interesting concept, along with feign's ideas, perhaps a mirror inside cube with the peachy generating several scan lines (and moving them) perhaps with the cameras in the upper corners of the cube, but the issues with this varient are significantly higher processing needs from the several cameras, but the benefits are reduced time to scan, with the need for few(less than 5) bursts of scan and then stitching the scans together along with the resolution of the camera ie higher res is approx higher detail/accuracy of the model

That diagram has a lot of the concept right, though in this case, the Peachy can theoretically get in its own way... However, looking at it diagrammed out, I see that the only way it can actually hit itself with the laser is if it misses the object many, many times. Also, in that configuration, the sweep of the laser has to be precisely aligned in both axes to form a coherent line. If the laser is above the box, then it can scan across the part, only worrying about being perpendicular to one pair of mirrors at a time.

I haven't built anything out yet, though I've done a little bit of work with crazy "infinite mirror" boxes before.

Quote:

---

Originally Posted by Feign

That diagram has a lot of the concept right, though in this case, the Peachy can theoretically get in its own way... However, looking at it diagrammed out, I see that the only way it can actually hit itself with the laser is if it misses the object many, many times. Also, in that configuration, the sweep of the laser has to be precisely aligned in both axes to form a coherent line. If the laser is above the box, then it can scan across the part, only worrying about being perpendicular to one pair of mirrors at a time.

I haven't built anything out yet, though I've done a little bit of work with crazy "infinite mirror" boxes before.

---

I've just ordered the bits for the sardauscan. All the .stl files are there, so printing these using Peachy should be relatively easy. It's a ready made 4 laser solution for about £12, as I already have an HD webcam. Not too shabby. That's not to say I'm not looking forward to the Peachy solution.... because I am !!!!

https://www.youtube.com/watch?v=TPb-gEVkBYo

Quote:

---

Originally Posted by rylangrayston

Great Ideas everyone!
Iv tried using a very old version of the peachy to move the laser in a line quick enough to simulate a line laser lens.. its a great idea and it still has potential.
one of the problems with doing that is the that typical web cams dont give you control over the exposure time, and the exposure time adjusts automatically, so its tricky to
get expose a line that dosent overlap uneven brightness. The first scans shown in the main campaign video did use a peachy printer with one mirror vibrating very quickly. and as a temporary solution I used a DSLR camera which allowed me to control the exposure times and minimize this problem.
I like the idea of a scanner hack where a line laser ( with a line laser lense) is placed in the peachy printer head and one of the mirrors causes the plane of light coming from the laser to move across the model.

We also wrote software quite some time a go that takes N video frames and mixes them together to create a long exposure shot... for example filming at 30 frames per second, you can take 60 frames to get a 2 second exposure image. This could be useful when drawing a grid.

I Think the first version of the scanner will be simpler than these ideas but I can see how as scanner hack today could become the standard 2.0 scanner of tomorrow.
oh and something to stop it a precise spot... yes Ill see if I can work that in today, seems useful.

---

Rylan.... I've done a lot of googling and thinking and I definitely think your idea where Peachy sweeps a vertical line laser across the object is definitely the way to go. This duplicates scanners with multiple lasers like the sardauscan. Version 1.0 of Peachy Scanner should definitely go for that and I think like sardauscan, it has a high chance of success. People could use the Peachy Printer to print any required support items.

What if we were to use a polar platform (a rotating platform off center of a rotating platform) and have the laser scan vertically, sweep horizontally, and then diagonally? Yes it would take a long time to scan but wouldn't the scans be more complete?