A neat photographic method for capturing 3d surface

[bill.young]

These techniques are getting better and better...

http://www.hothardware.com/News/Capturing-3D-Surfaces-Simply-With-a-Flash-Camera/

[handh]

That's pretty neat.

[nschlee]

Where can we get the software that compares the two images and generates the ht. map???? :-)

[dana_swift]

Makes me want to write something!

Combining those techniques, with the well known mathematics of optics and photogrammetry a very nice 3D rendering should be possible. If it just needs to "look good" for a carving, that would be a simple method of getting it.

Hmm this is sounding like fun!

D

[jseiler]

If you are looking for a way to get shallow relief from simple flash photography, check this out:

http://zarria.net/nrmphoto/nrmphoto.html

I did some playing around...

http://www.talkshopbot.com/forum/messages/312/24603.html?1192018777

The results can be spectacular. Its not a laser scan, but it can get you some pretty artistic shallow reliefs really fast (not particularly accurate, but very artistic). I wish I had time to experiment more, but I ran into a brick wall of not having the right software (GIMP vs photoshop) and not having the right project to pay for the investment in better software (necessity is the mother of investment?) Maybe someone with some spare time and photoshop would like to run with this and show us something really cool.

John

[ron_varela]

Here is something that has been around a few years and worked in our furniture restoring when we did antiques.

http://www.david-laserscanner.com/

Ron

[dana_swift]

Bill & John if you took the bumpmaps created with the technique used in the article at:

http://zarria.net/nrmphoto/nrmphoto.html

Then use Maple, Matlab, MathCad or some similar program to compute the blue (Z) component of each vector. Knowing that 1 = X^2 + Y^2 + Z^2. (I'm a Maple user, and John, I saw you have Matlab.)

You should be able to generate excellent normal maps. Converting those to 3D relief maps could be done using any of several methods.. perhaps the most simple for non-critical reliefs is to take the dot product of the corrected normal vector with the "k" unit vector in Z. I can think of others which would result in a more true relief.

The ideas are gaining refinement- the first step to serious inspiration.

D

[bill.young]

Dana,

Have you seen this pdf that explains the first photographic method in more detail...

http://aig.cs.man.ac.uk/publications/papers/daedalus/GlencrossWard2008.pdf

(it's all Greek to me!)

Bill

[nschlee]

Go for it! Dana :-)

Neal

[ron_varela]

Here are the results I made from the links provided by Bill and Dana.
I used the pictures of the peanuts and the method used by Bills link, this might not be for the average shopbotter.

Ron


9132

9133

[mitch_prest]

any chance of an explanation in english as to what you did to achieve these results..

linguistically challenged.. no greek..

mitch

[magic]

Ron

May I ask if the David system is good to capture people?

[ron_varela]

I will try to explain this as simple as I can but one needs to have an eye for detail.
First you find a good picture like this


9134

Then you need to make it a negative


9135

Then you need to make it 60 colors grayscale


9136

Then you need to use brightening and contrast to bring out the detail


9137

Then make it a negative again


9138

Then adjust the color to bring out the detail


9139

Final results


9140

The above method is new and took me over 7 years to perfect it by trial and error.

This is what I use to do 10 years ago before Lithophane came out.


9141


9142

The machine time to cut this 10x14 was only 10 minutes and I sold 100s of them for $25.00

The pictures were saved as a tif uncompressed and was scaled to 100x?...? means what ever the value was for the width.


9143

Now look at this I got over 220 shades of gray, the best is 256 shades of gray and that is what you want.
256 shades of gray will give you 25 levels of depth.


9144


Now to control the size of the picture you have to tell it the pixel size, so this was scaled down to 100x? Pixels so I gave it a value of .09 to make it a 9x8.


9145

My first test doing the picture above took only 10 minutes to cut and if I was to get the same results like this


9146

You need to scale it bigger like 1000x?, this will give you more lines and more time to cut.
Then you enter the maximum depth to -.25 for ¼ inch height or -.50 for ½ height and so on
This is like stopovers set at .010, just use a 1/8 or ¼ ball nose to get the results you want.

This is the best way I can explain the method in lemans terms.

Ron

[ron_varela]

David yes you can do faces, you will want to make sure they keep their eyes closed, then just use a model for the eyes


9147


9148

[magic]

I've been playing around in photoshop by making the image 30 shades in grey in a tif format, Then using the histograph to set a high and a low then opening it in AC and that's working pretty well. If not the histogram I'm adjusting the highs and lows by washing the entire image with a light grey.... but this is an artistic approach (it works) but i don't have any numbers... I should write down the C values, next time.

On a 5 X 7 I'm using a 1/8 bit with a 10% step-over and it's taking 30 minutes.

- it looks like you're getting great in the photo, with a 1/4 ced, and I can't see the step-over lines.

Is there a different geometry to the ball then round? is it flatter because it seems that you're cutting more volume in less time.

[gerryv]

Just now came across this info so I'll cut & paste it here as is. I'm a Mac user so sorry I can't test this out. Seems right along the line of this discussion however. Here goes:

Microsoft has publicly released Photosynth, a way of combining conventional images to create 3D scenes. After you upload a set of images, the software analyzes each for similarities to the others, and then it uses this data to build a model of where the photos were taken. A viewer can then browse through the final photograph, navigating smoothly and zooming in tiny details.

What's the idea?

Humans are able to perceive depth by instinctively calculating the effect on perspective of the offset between their eyes. Conventional digital cameras, with a single viewpoint on the world, cannot do this. Photosynth identifies common features in multiple photographs and uses them to work out how the images relate to one another. It then uses this information to build up a 3D map of how the features in the image, and the positions of the cameras that took them, relate to one another.

The software can combine images shot with the creation of 'Synths' in mind or by mixing images taken at different times, dates and resolutions.

It is the first use of Microsoft's much-hyped and rather astonishingly-named 'Seadragon' technology. The accompanying Photosynth blog provides some of the background. The team has provided video and pdf instructions for creating your own 'Synths,' including a guide to subjects and photographs that will be considered 'Synthy' and those considered 'Not Synthy.' (It turns out that Venice is considered distinctive and feature-full enough to be Synthy, while the Seattle Public Library remains stubbornly resistant to Synthing).

[Brady Watson]

The subject of digitizing 3D objects is as diverse as the subject of cutting objects on a CNC. You must first ask yourself what it is exactly that you want to achieve when creating or acquiring 3D content. In my experience, there is no one silver bullet that is going to meet every need in your business, and more specifically, meet every need of your customers.

The examples shown above are already greyscale, and ArtCAM can easily make a relief from those types of pictures - and a greyscale from a relief. It seems that someone has already done the work, all you need to do is convert a greyscale to a relief that represents XYZ. Finding good greyscale pics is difficult at best. Regardless of the method used, creating a 3D relief from a photo often results in disappointment, as most photos simply do not have enough information in them to create much more than a very low relief texture, which can be applied over top of an existing model. Bill Young's post of using 2 images, one with flash, one with out, is promising for creating certain types of decorative reliefs. How accurate it does this is yet to be known - and that brings up the subject of some digitizing technologies not really being suitable for anything other than decorations. You certainly aren't going to digitize a part with critical dimensions this way with just 2 pics...

Photogrammetry is another spin off, and that can be used in the 3-axis CNC world mostly for 2D tracing & digitizing. You can also take side shots, create a vector cross-section from them and do rail sweeps when the model you wish to create is conducive to that process. Programs that use photogrammetry (like PhotoSynth, PhotoModeller etc) are really geared towards large objects that do not have a lot of curves or embellishment. They are of little to use to most botters. So...you're back to finding something that really fits your needs.

Scanners...probes...lasers...all sound sexy, but keep in mind, they also have their caveats as well. Touch probes often gouge parts, often result in noisy scans and sometimes are less than reliable. The SB probe is excellent for most models over let's say 8" square or above with no more detail than a 1/16" ball end mill could carve. The perfect companion for the SB probe is the Roland Picza. It will scan everything below about 8 X 12" at step resolutions up to 0.002" - It is a very impressive tool that pays for itself quickly if the reason for buying is actual work...not just because ya wanna have one...which is true for all of this stuff - and ya get what you pay for when it comes to investing in digitizing equipment. For most botters that want to get into digitizing 3D, the SB probe, a used Roland Picza, good software that can sculpt, and your brain are all that is needed to go very, very far in the land of 3D - without disappointing yourself or your customers.

Lasers...3D cameras...Microscribe/Faro arms...the pinnacle of digitizing...or so it seems on the surface, per se
Lasers have their problems too. Color transitions, shininess, linearity of the sensor & sensor range all come into play when you talk about lasers. 99.9% of the lasers out there have a max grazing wall angle of about 60° on a good day. This means that when the laser meets a vertical plane, the laser cannot see the laser spot, and in turn doesn't report a Z-value...so you get drop out - portions of your scanned relief that are missing. The NextEngine does this consistently on flat, decorative reliefs and does it in areas that you wouldn't expect it to. 3D cameras do the same thing - lots of drop out...and the resulting 3D content is often very noisy. It is one thing to smooth & work on a relief that is to be cut on one side via a 3-axis machine, but an entirely different animal when you have to sculpt in a full 360° - even if you can unwrap it. Microscribe & Faro arms are great for large models, most often too large to cut on the bot, and the process of using the arm takes LOTS of time & coordination to execute properly. The DAVID scanner is kinda in it's own class as far as digitizers go. It is extremely sensitive to various conditions, and is most suited, in my personal experience & opinion, as a weekend science project. Even with excellent optical equipment & a clean laser source, the resulting 3D is disappointing. You can make out some features of the model, but it isn't at all what you expected. Last I saw (could be different now) the DAVID site showed a decent cherub scan & some guy's face. That's about it. I expected better considering it has been brewing now for a couple of years. I was developing my laser setup when the DAVID came on the scene...I really wished to all get out that it was as good as the creator of the DAVID originally hoped it would be.

Faces...There are a few ways to get faces into your computer. First, you could invest in a 3D camera and have a full scan of someone's face. The investment is around $15-20k. Plus software. Two, you could use ArtCAM with the Face Wizard tool. Next, you could use a program called FaceGen, which creates a hairless model from 2-pics - you get to sculpt in the hair in another software package. Finally, you can take a mold of someone's face (if they are willing & alive) - and then cast the impression with plaster or plastic. Then probe the result on the bot or on your Picza.

It all boils down to what you need. Obviously professionals have very different needs than hobbiests...so choose your equipment & method(s) wisely.

-B

PS - Rather long-winded, I know...but this should help people that want to know this sort of stuff. I get a LOT of emails asking about what I've written here.

[ron_varela]

David is a nice program to have and the fact it did a nice job on the face and the cherub says it all.
Just because no one has uploaded any of their work does not mean it can not do the job and best of all it next to free.
I have used it many of times and worked for me and did the job for us, and did not have to pay someone to make the files for us.

From what I read on many forums Artcam can not do a frontal face only a profile and not all that good, the face wizard has been beaten down by experts and gurus.
Talk by Artcam users is the sample they provide of the woman is the only picture that came out fairly well and no one has produced anything close to it without major cleanup or modifying the picture, the trick is LIGHTING and how to place it, black is always the deepest but using the tiff converter you can tell it if you want white to be deepest, same as inverting.
To make 3d models like this http://www.3dmodelclub.com/index.php you will need to buy a draw pad “$800.00 and up” along with a good graphic program like Photoshop or similar.
This is the picture I had


9149

And turned it into this


9150

All I did to it was remove the noise and add a light grayscale brush and added 12 lights to it
The trick behind pictures is the information in the picture, so if a picture does not have enough information give it some.
I have been playing around with grayscales for the last 12 years and know quite a few tricks on how to get information in a picture, one trick is to clean up a picture then print it then scan it with high resolution and get the file around 3mb or better, the files I do are around 50to 125mb to get the detail I want then resize it, have a good scanner too.
You can get many 3d programs for free like Bryce – zbrush - Bishop3d – Blender - Faceworx creates a 3D head - wings 3d and there is information on how to use it.
The bottom line is you will need to learn how it all works, there is no one 3d program that is the Holy Grail and there might never bee one.
Like I said above “This might not be for the average shopbotter”
My 2 cents worth is get the free stuff and play with it and ask questions on their forum, or do what everyone in business like I was have the work done for you, you save time and money.
Get Brady to make you the files or others

Just 1 of many Scanning Services
http://www.100megs22.com/~clipart/index.php?main_page=product_info&products_id=2&zen id=07c58cc57e5e9176e927368636817d99

Ron

[dubliner]

Is it my computer or do the 100 megs files look dreadful, the animals are disproportionate, very rough & one of the horses actually has a breast attached. I dont think James Booth has anything to worry about if thats his competition. Has anyone carved one of these or God forbid, paid for one? They look like something I would make :-)

[normand]

I feel good today
http://david-laserscanner.com/forum/viewtopic.php?p=3672#3672
Go slowly to the bottom of the page
In David forum you have to look at the gallery the (user gallery)
Some people are braggy what on their mind I dont know ,to aim the light on them and their shadow on the thrue?
Funny Ron couple weeks ago you e-mail me wanting to sell me a cd of 3d file .I did not need but feel sorry for you with all that happend and gave you that link ,you have never know of it. Now you say you been doing antique ?
Little humble pie never hurt
No need to answer I know I wont respond
PS Blender is free ZBrush is not
Normand Blais

[ron_varela]

Normand I think you misread what I said, David has been out for a few years and the link was posted here on this forum, so if you search this forum and the archives a few years back you will see it, many shopbotters knew of this site and downloaded the software but like all new software there was very little documentation listed by the people that used it and it was all trial and error, the software did work well for small wood items so we used it for that before I retired.
I have been a member of this forum for 10+ years as with Bill Palumbo and Bill Young, they are the only 2 I know still here posting great information. I have made thousands of post but can not say it was me that posted the link.
Like any forum that has 1000s of post it is simple to have post buried and looses its interest.
Just like this topic will be buried in time and nothing will be added to it just read by who see it.
If you did email me the link then I must have went and took a look and seen the new updated site and felt it had more helpful information shopbotters might want to try it out for themselves, all this would have been was a revision of what was old and now what is new so I posted the link in a topic that was already made. This gives people a new chance to go try it their self and ask questions on what is best to use.
I do not take any credit for having the link you might have sent me, I just posted I have used it in the past and it worked for my needs.
The link you just provided is a great asset to the software because everyone can see what it can do http://david-laserscanner.com/forum/viewtopic.php?p=3672#3672
I have read the email you sent me a few weeks back and I thank you and all the other shopbotters that did the same, I took a hard loss and will get over it, my only love now is to continue to what I do best, help everyone the best I can.
With my skill and knowledge it would be a shame not to share it with everyone. With my loss and my passion for being able to have a shopbot will always be close to my heart.
I might be old and sometimes leap before I look but I am still learning, with many chat rooms and forum it is always out with the old and in with the new so it might be time for this old shopbotter to fade away and be buried in the archives.

Ron

[dana_swift]

Follow up- I tried an experiment based on the paper from the link at the top of this thread by Bill Young (thanks Bill!). First I photographed a figurine with a flash as the paper suggested:


9151

This shows the 3D figurine has various colors, and reflectance values as well as 3D texture.

Then I lit the figurine from the side with a regular table lamp. Using the flash image to remove the effects of the figurine reflectance and color (albido) I created a "depth image".


9152

This image clearly shows the texture of the figurine with the color and reflectance removed. While the technique is promising, it has far too much error to continue processing the image into an XYZ point cloud. If this image had been much more "evenly gray" the process would have been good enough to continue writing a full 3D conversion program.

I can see the technique as being useful for capturing textures, at this moment it is discouraging as to recovering the full volume scan. But I'm glad I ran the experiment!

I'm not entirely shut down, but a technique needs to be devised that more accurately removes the albido effects. I can think of some ideas that might be of use, but that will have to wait for another day.

This was done in short C# program, using the Microsoft provided image libraries. I won't bother explaining the math I used, it is basically what is shown in the paper with a little of my own "sauce" added..

If anyone cares to know what I learned- there it is!

D