Methods To Sense The 3D Surface/Structure Of A Book

[Anonymous1]

And with a little GIMP magic, I corrected for the lens distortion beforehand, so the results look even better than before:

Correction for radial distortion helps a lot.

canvas.png (691.3 KiB) Viewed 563 times

That keystoning on the top is due to the camera's position. It's not perfectly centered (that's basically impossible to do), so I'll have to live with it. I think BSW could correct these really quickly, as I am not going to make a complete remake of ST in Python (although I have made a page splitter with OpenCV, which works fairly well).

[steve1066d]

Very cool.. I'm in the process of getting OpenCV installed on my linux system to give it a spin.

[steve1066d]

Anonymous,

I've taken a good look at your code, and I like your approach (as much as I can figure out anyways).

Could you add some comments to your code? I'm going to try to move some of that into java, but since I don't know much about OpenCV, Python or the math involved, I'm at a bit of a disadvantage as I try to follow it

[Anonymous1]

I've added tons of comments. My commit should show up in a minute or so.

[Gerard]

hi maybe a photo slide projector http://www.instructables.com/id/Photo-s ... patches-t/ could be more flexible way to project lines, qr-codes or grids on the book
or an mini led beamer

[U235]

Hi everybody.
I have easy idea 3D surface reconstruction. No lasers.

P.S.
Sorry, i've got some problems with my English.

[daniel_reetz]

Hey U235, that is a cool idea. It resembles an idea that Spamsickle proposed before, except that he suggested to have two cameras looking at the edge of the book. With your method, you would only need one camera, and the mirrors could fold flat.

I think the same software that we are writing for the laser method could be adapted to this method.

Thanks for sharing your idea. Don't worry about your English here.

[daniel_reetz]

Also, Gerard, I bought one of those laser grid projectors you mentioned earlier... I'll post my results with it soon, I hope.

[steve1066d]

Anonymous,

Thanks for the comments. It definitely helped clarify a few things.

I'm going to combine our two approaches. I'm going to use the 3d model approach that you are using, but do it backwards. By converting the source image to a bunch of small plane segments, then determining which plane segment intersects the line that passes through the camera to the destination pixel, I should be able to get the source location accurately.

Using the source warp has advantage that you can use interpolation to keep things smooth.

FYI, here's my existing code. I probably won't have time to work on this for a couple of days:
https://bookscanwizard.svn.sourceforge.net/svnroot/bookscanwizard/src/net/sourceforge/bookscanwizard/dewarp/LaserUnwarp.java

[daniel_reetz]

Atarkri, Steve, Anonymous,

I mailed your systems this morning. Unfortunately, I had to use the real post office instead of my shipping system because my laptop died. I hate the post office! The cost of shipping was almost as much as building the things themselves, and they are not "guaranteed" to arrive until 3/1. I'm only complaining because my own shipping system apparently has a steep discount relative to shipping in the office itself. I apologize for taking so long in getting these things out.

That said, I'm glad that there's some hardware winging your way and I hope it works well. Steve and Atarkri's units are CHDK compatible, and Anonymous's has a 3.5mm stereo jack (and cable) with the ground and tip connected to the second switch. All are powered purely on AC and don't need any batteries.