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

[steve1066d]

I think my code is actually working.. Its not perfect.. but its definitely getting there. Here is a dewarped image:

Here is the image cropped and keystone corrected. Fixing the keystone error might be considered cheating, but I think the same correction will apply to all the pages in a book, so I think it is a valid way of handling things that aren't quite lined up correctly. Once I try a whole book we'll know for sure.

[Anonymous1]

That looks amazing, Steve! So are you just straightening the laser lines (like in the test image with a grid)? I'm a bit confused...

But this is amazing work.

[daniel_reetz]

amazing, Steve!!

[rob]

Well done!!

[Shaknum]

Great work! That is fantastic.

[daniel_reetz]

Hey everyone, I need some discussion of how to put this thing together. I have the electrical side neatly figured out, but I want to make sure it's a good experience for the operator, too.

The basic setup is a little counter-intuitive.

1. The camera must be "portrait" along with the book, to capture two pages.
2. The lasers lines must be parallel with the top and bottom of the book, meaning that
3. One of the lasers "wants" to be where my head would normally be when scanning a book.

Here's an illustration:

Now, I'm thinking about a couple of things. One, the system COULD be just an attachment to a pocket camera, on little "extensions" that pull out, like this "pick up tool"

pick-up tool.jpg (142.78 KiB) Viewed 6429 times

- in this version, it could be portable and go into a messenger bag w/o problem.

Two, the system could be two stands that rest on the table, plus some method of supporting the camera. So you just set up two little towers, align the lasers on the table surface, and position your camera. The problem becomes camera support, how is the camera held up? This is less of a problem in the case of light and small cameras like Steve's, and more of a problem for SLRs like Anonymous's.

Three, the system could be some kind of "overhead truss".

Or something else.... ideas?? I plan to spend the night on these, and although something could stop me, I think I have as much time as I need. This morning I'm going to go buy parts from All Electronics and get things going...

[TomHorsley]

I've been wondering if using two cameras could provide more accurate 3D data from the stereo image (no actual idea of how it might help, mind you - just wondering).

[steve1066d]

Here's some initial thoughts:

Portability not a big concern for me for a prototype system.

It's helpful to adjust the laser angles. I think it would be easy enough for the software to support other angles, and it would be good to experiment with different angles.

Awkward page turning isn't too big of a deal at this point.. I'd rather have something flexible.

I will want to experiment with camera at different heights, as I think this could help limit some distortion in the first place. I think the truss idea would restrict the height to the height of the truss.

The telescoping idea I would think would be too sensitive to vibrations.

I'll do some more thinking about it this afternoon.

[daniel_reetz]

Alright, I hear that loud and clear - I'll make it adjustable and flexible over portable/finished. That's easy enough to do.

I can make the available angles "detents" - such that you have 14,30,45,90 degrees - or I can make them continuously adjustable. I can possibly do both. Preference there?

Adjustable height and width of lasers will be part of the design.

I bought supplies this morning, and I am going to be including a lighting system for you, too. I have enough to make four systems right now, and that's probably all I can realistically build until I get back from traveling, so it looks like Atarkri, Anonymous, and Steve are going to get this first batch, leaving one for me. If someone else wants to get involved in software development, I could be convinced to produce more systems in the near future, so feel free to contact me about that.

[steve1066d]

I think the detents are fine. I'd imagine we'd normally want to stick to standard angles, and I'm sure I can jury rig something if I want to use a weird angle to test with.