The "Nabu" - an automated book scanner

[Morten Hansen]

In this thread I would like to share my work on my automated book scanner, nicknamed "Nabu", which I have been working on ever since I got my Rep 2 some six years ago. The design is currently in its 10th iteration. A LOT of trial and error has occured. 

The idea

I have a book collection of about 800 books, mostly about history and about 75% them are in Danish. So buying them as audio books is not possible. And I want them as audio books,because since I have very little spare time to really read books, listening to them during my commute to work is the second best option. Now, I'm not going to manually scan 800 books (though there are some awesome manual DIY book scanners designs out there), so that's where the Nabu comes in. The pages will be cleaned in ImageMagick, OCR-processed with Tessaract and put through Googles WaveNet text-to-speech service. And hopefully I will have some kind of home made audio book. There are A LOT of issues with these steps, each one is a minefield of its own, but I'm confident I can get something workable out of it

The scanner

The idea is to put the book in a motorized cradle in an 80 degree angle; that cradle will move as more and more pages are scanned, making sure the center of the book is always in in the right place for the cameras. 

A vacuum device will fix a page, turn it over, and a camera will take a picture. The vacuum device will then be lifted and a thin "finger" will turn the page. The vacuum device goes down again, sucks up the other side of the page it just let go of, turn that, and a picture is then taken of that side of the page. Then a lift and a final page turn. Repeat until the last page. The motor control is done using a duet wifi board, for which I'm making an Electron app. For the cameras, I'm using Android mobile phones with 40 MP cameras, equipped with a simple app that I've made (but which needs some further development).

So far the cradle is done, which was easy, because it's basically the same design I used in my other builds. It has Nema motor with a lead screw for movement, and a lead screw with a knob for manual adjustment of the spine distance. There are a number of improvements I want to add, but that can wait for version 2.0.

The problem is the vacuum device and the page turning unit. That's what I will be focusing on in the coming months. And something I hope the good people here can help with.

[Morten Hansen]

<b>Placeholder entry</b>

[Morten Hansen]

So yesterday I finished the construction of the first prototype of the vacuum device. Unfortunately, I had a feeling that the diameter of the air flow pipe was way too small (diameter 10mm) and I was right. It was plugged into my vacuum cleaner through a 3D printed adapter and while it did "suck" the pages in place, it required the vacuum cleaner to be turned up to its maximum level. A previous model I made las year didn't require that much sucktion, but that prototype had a 32mm diameter air pipe input. So, back to the drawing board to add more air flow.

My first prototype of size A3 vacuum device. Air flow input on the lower right, with a silicone tube

The device before the peghole sheet was glued in place. 

Every hole was hand drilled and filed for smoothness! I'm not doing that again

[Kurt at VR-FX]

Am curious about this posting Entry Morten. Is it just to separate the 1st & last posting entry? I guess if you read the forum via the Google webpage - it maybe makes sense. Although I purely read via email. Anyway - was just curious...

-K-

On 6/14/2020 12:17 PM, Morten Hansen wrote:

Placeholder entry

--
You received this message because you are subscribed to the Google Groups "3D Printing Tips and Tricks" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 3d-printing-tips--...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/3d-printing-tips--tricks/023abad5-b128-4036-96af-c9461143bba7o%40googlegroups.com.

[Morten Hansen]

Yep, I'd like to have a post or two with some content that can be updated, so that a reader can go to the start of a thread and get the latest info there, instead of having go through every single post. The design of this machine will evolve over time, so if anyone would want to see the latest iteration, they could go to the start of the thread.

søndag den 14. juni 2020 kl. 21.56.57 UTC+2 skrev vrfx:

Am curious about this posting Entry Morten. Is it just to separate the 1st & last posting entry? I guess if you read the forum via the Google webpage - it maybe makes sense. Although I purely read via email. Anyway - was just curious...

-K-

On 6/14/2020 12:17 PM, Morten Hansen wrote:

Placeholder entry

--
You received this message because you are subscribed to the Google Groups "3D Printing Tips and Tricks" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 3d-printing-tips--tricks+unsub...@googlegroups.com.

[Petr Ptacek]

Hey Morten, I'm still not sure how the process of flipping pages with suction device works. Could you elaborate a bit on that?

[Morten Hansen]

mandag den 15. juni 2020 kl. 23.13.09 UTC+2 skrev Petr Ptacek:

Hey Morten, I'm still not sure how the process of flipping pages with suction device works. Could you elaborate a bit on that

Hi Petr, sure I'll try. I have a drawing that I hope explains it better:

1 - the suction device goes down

2 - the suction device flips over and "sucks" the page

3 - the suction device flips vertically. Now, once in while the page beneath the affixed one might follow. It's rare, but it happens, and that's why...

4 - the page turner arm goes vertically (a view of it from the front can be seen in the lower right corner)

5 - the "finger" of the page turner slides down just next to the fixed page

6 - any "rogue" page will now be turned back by the page turner

7 - the suction device flips over to its scanning position and the camera takes a picture

8 - the suction device turns back to the vertical position and goes up, leaving a gap of about 1/3 to 1/2 og the page size

9 - the page turner turns and goes beneath the gap, (the suction device has stopped "sucking"), and the page turner takes the page with it

10 - the page turner moves itself out of the "work space" zone and the whole process is now repeated for the other side of the book.

When the mirror process is done, the page turner flips back to its starting position to the right, and we start all over.

[Nick Hall]

Hi Morten, how does the vacuum device come down into the work space? Just trying to get a clear picture in my mind. I am also thinking that it doesn't need to be as big. What if you had another arm, that spanned the work area and just had a suction tube that would grab the top of the page and then rotate the arm across the work area, to position your page as necessary. Turn off suction, lift arm out of work area and repeat? Should be able to get more suction with the lower power from a smaller vacuum area.

[Morten Hansen]

I have a screenshot of the Sketchup model I'm doing for the next version, hopefully it gives a better sense of the design:

I've considered the same idea that you had, but I ultimately went with the big suction board for a couple of reasons... it lays the entire page out completely flat, which gives the camera a better chance of making a good scan with straight lines. A skewed scan could be corrected with some software magic, but I'd rather avoid that. A large opaque suction surface also avoids the issue of background lighting, meaning light that goes through the paper and makes the letters on the back side of the page visible. I did a test last year with a simple device that just held up the paper and then I photographed it. It works, but it doesnt take a lot of light for unwanted artifacts to show up on the scan.

[Morten Hansen]

Nick's comment gave me some ideas how to use only a few small suction areas, while still having an entirely flat, smooth surface. I'll post the work on that in a couple of days. 

In the mean time I've been experimenting with some of the old concepts for vacuum suction I did last year, updating them a little (image 1-5). It works really well, and maybe a little too well. It seems like the suction goes through the top page and also sucks up the underlying pages. Which makes sense, given that paper isn't necessaily a solid mass. It's something I need to take into consideration.

The last image is the full board prototype in comparison, but this time hooked up to a better vacuum cleaner (our old vaccum cleaner burned out a couple of days ago when I ran a longer test on the full board! The machine was 10 years old, but I still think it shows that if a vacuum cleaner is needed to run during the whole scan process, it probably shouldn't run on maximum power... hence why I'm working on fewer, smaller suction zones) 

[newmo...@gmail.com]

Summer came and went and I was very close to giving up on the project. I have tried every kind of vacuum solution that I could think of, but I always came to one of two results: either the suction was to weak and the device would not lift the page reliably; or the suction was too high, and the pages beneath the upper page would also be lifted up... as mentioned, paper isn't as solid as I thought (or hoped), so air travels through it. The flow of air would have to be very, very closely controlled, in a way that would require very sophisticated controls, and it would have to be set relative to the size and weight of the page it was targeting.

Another problem was that if suction continued for too long, my vacuum cleaner would overheat and turn off. I believe the vacuum cleaner uses the through-flow of air to cool its motor, and when the paper was beeing sucked, there wasn't any flow of cooling air. So I tried to make a venturii system, but that only took care of the overheating of the vacuum cleaner. It didn't fix the problem of over- or undersuction of the paper.

So I was about to give up and move on to making a manual book scanner, but then I came across a research paper that featured a socalled nearfield electroadhesive device to lift a single page pretty reliably for an automated bookscanner.

https://osf.io/gq8ta/download

Problem was, they made the electroadhesive pad themselves, and while I found some recipes for such a device on hackaday.com, I was not super excited to make one myself. I have very little experience with voltage converters, silicone molding and that sort of stuff. The technology is pretty common in an industrial setting, but buying something smaller and simpler was turning out to be impossible.

Enter a rather silly office supply product, the Justick Desktop. It's basically a sort of bulletin board for your desk, that uses nearfield electroadhesion to make paper and other flat items stick to it without use of pins or magnets or tape: https://www.justick.com/products/. The desktop versions surface area is around the A4 mark, so that was perfect for my use. I got one, tested it, and it works perfectly! The field is strong enough to lift an A3 page with little difficulty, but the field doesn't travel through the paper in a way that catches the page beneath it. Heureka!! Just to be safe, I bought 2 more, because it seems they are sold out, at least in Europe.

So here I am, 4 weeks after I found the solution, making an arm, that will hold the Justick Desktop, and use that for lifting pages just enough for the paper flipping finger to get underneath the lifted page and do its thing.

[newmo...@gmail.com]

(Edit: re-post... apparantly you can't embed a youtube video here?)

So, the Justick Desktop idea didn't really work... It was only about 75% reliable, and that's just not gonna cut it. So I tried a bunch of other ideas, including using suction cups after it was recommended to me as an industry standard way of handling paper. However, that gave the same problems as the vacuum solution: it sometimes grabbed the page beneath the primary one, and that won't do. However, I noticed that the rubber cups have great grip, so I made the following:

Youtube link: https://www.youtube.com/embed/9weF24AeUR8