Full Auto Brainstorming

DIY Book Scanner Skunk Works. Share your crazy ideas and novel approaches. Home of the "3D structure of a book" thread.

Moderator: peterZ

User avatar
jck57
Posts: 376
Joined: 23 Nov 2009, 15:21

Full Auto Brainstorming

Post by jck57 »

Join the brainstorm. Comments welcome.


The hardest part is to consistently pick a single page. This is because of the great variations in page quality, book size, and bindings. Any dependable machine must be equipped with page fault detectors to reject multiple pages and nil pages. A desirable feature connected to page fault detection is smart control that can alter page-picker adjustments to avoid future faults on the same book. Once we get to these smart controls we make possible a fresh approach to page picking itself. Instead of trying to carefully set up a delicate page picking mechanism, we go with a scattergun approach. We have a random attack on the page stack and then when successful in picking a single page, the smart controls catch that page and the page-flipping process continues. The random attack could be compressed air shooting in various directions at various ports at various velocities with or without mechanical fingers that attempt to separate pages. A possible way could be a finger equipped with compressed air jets. The finger approaches the page stack from the side, shooting air at the stack. Air jets are arranged so that there is no way the finger could physically hit a page or pages on end. The air blast coming out of the point of the finger will always move the paper away from the point. As the finger enters the book from the side the page fault detector checks the page and if a single page is detected, the finger is mechanically lowered to the book to keep the other pages in place. Next, compressed air jets on the finger flip the single page to the left-hand stack. If the detector shows a fault, the process is repeated with another random attack until something works. There are many variables available so even if the pages vary within the book, say because of glossy photo pages, the randomizer will find a way to separate a single page every time except when pages are virtually glued together.
User avatar
daniel_reetz
Posts: 2812
Joined: 03 Jun 2009, 13:56
E-book readers owned: Used to have a PRS-500
Number of books owned: 600
Country: United States
Contact:

Re: Full Auto Brainstorming

Post by daniel_reetz »

Here's a thought. What about semi-auto turning? As in, the scanner can ALWAYS detect if a page was turned somehow. So if it fails, it simply asks a human for help...
dpc
Posts: 379
Joined: 01 Apr 2011, 18:05
Number of books owned: 0
Location: Issaquah, WA

Re: Full Auto Brainstorming

Post by dpc »

Daniel, if we're talking about going "full auto" then any human intervention is forbidden. It's an all or nothing contract. It's the sort of process that would take place inside a hypothetical "scan-it!" kiosk. A book goes into a slot and out comes a DVD with the scanned images/epub/pdf/etc..

So if that's what we're brainstorming, you absolutely need a reliable feedback mechanism composed of a way to determine if a multiple pages were turned and then some way to back up and change the page turning method.

Jack, I don't know if "random" is the best term for what you resort to in order to separate multiple turned pages. I would imagine that you'd have a number of known fall back methods that would be attempted in succession rather than randomly. I say this because some of the fallback methods may take more time or put more strain on the pages than others and you'd want to try the relatively cheap & easy methods first. It could be that after having to resort to one of the more thorough fallback methods a dozen or more times, the process would just continue to use that method for the rest of the book as the time to try the earlier methods that repeatedly don't work could actually take more time than using the more thorough method from the git-go on each remaining page.

Regarding the use of air jets to blow pages apart... I still believe the most effective way to turn a page is a "pull & curl" of the page. The pull (typically using a vacuum or mild adhesive finger) could have a tendency to lift multiple pages (just as the other page turning methods do) that's why you need the pulling finger to be able to impart a curl on the page as it lifts the page(s). The curl puts a force on the pages underneath forcing then down to the stack of unscanned pages and helps separate them from the top page. Once the curled top page gets high enough, a wedge can move in from the book's side edge and push the page over. I'd imagine that the single-page sensing could be implemented on the pulling/curling finger and/or the wedge. If multiple pages are lifted, the mechanism imparts more curl to the page (i.e. curling tighter like the page wrapping around a dowel) putting more force on the underlying pages forcing them back on the unscanned stack and possibly turning the page a bit more slowly and changing the vacuum amount. You could always add the air jet separator to this rig as another fallback.

Anyway, that's how I would attack the fully-auto scan problem. I had actually begun design on just such a scanner a few years ago but then realized that the image acquisition part of the process isn't where I needed to be focusing my effort as I began to scan my home library. We're sorely lacking in the fully-automated post-processing side of the book scanning process. For instance, anyone can scan a 300 pg. book in under an hour manually with a simple copy stand (15 mins with a proper scanner), but it takes many hours to process the pages into something one can give to Acrobat to generate a PDF. That in itself wouldn't be so bad if the process didn't require a lot of manual intervention along the way. So these days I've been spending my time thinking more about automating the software side of things rather than being able to acquire page images automatically.
User avatar
jck57
Posts: 376
Joined: 23 Nov 2009, 15:21

Re: Full Auto Brainstorming

Post by jck57 »

dpc wrote:Daniel, if we're talking about going "full auto" then any human intervention is forbidden. It's an all or nothing contract. It's the sort of process that would take place inside a hypothetical "scan-it!" kiosk. A book goes into a slot and out comes a DVD with the scanned images/epub/pdf/etc..
Yeah, that's where I'm at. Dan's suggestion is a short term improvement but the goal is no human intervention.
dpc wrote:So if that's what we're brainstorming, you absolutely need a reliable feedback mechanism composed of a way to determine if a multiple pages were turned and then some way to back up and change the page turning method.

Jack, I don't know if "random" is the best term for what you resort to in order to separate multiple turned pages. I would imagine that you'd have a number of known fall back methods that would be attempted in succession rather than randomly.
Yes and well said. Through experimentation, the "toolbox" of page separating gets stocked with a number of useful "tools". As you say, if one doesn't work, the next one is resorted to until something works. The machine thus "learns" and thereafter applies the best tool first, only resorting to other tools if needed.
dpc wrote: I say this because some of the fallback methods may take more time or put more strain on the pages than others and you'd want to try the relatively cheap & easy methods first. It could be that after having to resort to one of the more thorough fallback methods a dozen or more times, the process would just continue to use that method for the rest of the book as the time to try the earlier methods that repeatedly don't work could actually take more time than using the more thorough method from the git-go on each remaining page.

Regarding the use of air jets to blow pages apart... I still believe the most effective way to turn a page is a "pull & curl" of the page. The pull (typically using a vacuum or mild adhesive finger) could have a tendency to lift multiple pages (just as the other page turning methods do) that's why you need the pulling finger to be able to impart a curl on the page as it lifts the page(s). The curl puts a force on the pages underneath forcing then down to the stack of unscanned pages and helps separate them from the top page. Once the curled top page gets high enough, a wedge can move in from the book's side edge and push the page over. I'd imagine that the single-page sensing could be implemented on the pulling/curling finger and/or the wedge. If multiple pages are lifted, the mechanism imparts more curl to the page (i.e. curling tighter like the page wrapping around a dowel) putting more force on the underlying pages forcing them back on the unscanned stack and possibly turning the page a bit more slowly and changing the vacuum amount. You could always add the air jet separator to this rig as another fallback.
I've pondered some on this idea since you first mentioned it some time ago. I'm going to do a mock-up and see how well it works. If I understand you correctly, a pipe is lowered to the corner of the page. This pipe has a slit that is covered by the paper of the page. Vacuum is applied to the slit and the pipe rolls away from the corner, pulling up the page as it goes.
dpc
Posts: 379
Joined: 01 Apr 2011, 18:05
Number of books owned: 0
Location: Issaquah, WA

Re: Full Auto Brainstorming

Post by dpc »

jck wrote: I've pondered some on this idea since you first mentioned it some time ago. I'm going to do a mock-up and see how well it works. If I understand you correctly, a pipe is lowered to the corner of the page. This pipe has a slit that is covered by the paper of the page. Vacuum is applied to the slit and the pipe rolls away from the corner, pulling up the page as it goes.
Yeah, that's the basic idea. I was thinking more about the pipe laying across the page vertically, then rolling back the entire page edge toward the gutter slightly as it moves up from the stack of unscanned pages enough for the wedge to come in from the side and push the page over, although rolling from the corner should have a similar effect.

If you want to see something like this in action, look at this video: http://www.youtube.com/watch?v=FIYlT0IvksU. That scanner doesn't seem to curl the page to push the underlying pages down but I think that would be a natural extension over what they've done to attempt to handle sticky pages. I also like that the platen glass is attached to the travelling wedge so that it keeps books that are naturally closed (i.e. paperbacks) open during the scanning process.

I've seen posts here where people seem to be against using vacuum to lift pages because they claim that the vacuum somehow goes through the top page and lifts pages below as well. I find that hard to believe myself, but you should be able to control the amount of vacuum so that wouldn't happen. The page curl would also prevent this from happening. When I was playing around with this, I just put some double-stick tape on a dowel and experimented with various page types/sizes and amount of curl by hand. It seemed to work well but I didn't have sample books with sticky pages to test.
User avatar
jck57
Posts: 376
Joined: 23 Nov 2009, 15:21

Re: Full Auto Brainstorming

Post by jck57 »

Idea for a page lifter with automatically resettable downforce. The P/E transducer is a pneumatic pressure regulator that is controlled electrically so the Arduino can put more or less force on the rubber tipped page lifter. If the page fault detector sees more than one page, pressure is reduced by the transducer. If no pages are lifted, pressure is increased. A glass-lined, low stiction Airpel unit would be a great choice for the spring-return cylinder.


pe lifter.jpeg
pe lifter.jpeg (36.45 KiB) Viewed 15484 times
dpc: In my own experiments even moderate vacuum on a page tended to pull on the page underneath.
dpc
Posts: 379
Joined: 01 Apr 2011, 18:05
Number of books owned: 0
Location: Issaquah, WA

Re: Full Auto Brainstorming

Post by dpc »

jck wrote: dpc: In my own experiments even moderate vacuum on a page tended to pull on the page underneath.
Interesting. Makes one wonder how that scanner I posted a link to is able to do it so well with vacuum? They claim that they can vary the vacuum based on the page detector sensor though. Maybe that helps? Thanks for giving it a try!

Update:
This morning I tried this using my Hoover connected to a small piece of scrap 1/4" copper tubing with 1/8" holes drilled along the length about 1/2" apart. It worked like a champ. I tried a rather heavy-paged Tom Clancy novel as well as an old Intel microprocessor manual (very thin pages) and never had a problem turning single pages using a combination of lift and turn to impart a curl on the page. I did find it works better if you place the lifting "wand" near the page's edge as it comes down on the page and then curl it slightly before lifting.

If you try to lift too far away from the edge you can create a vacuum under the page and cause underlying page to come up with it. This isn't the vacuum from the wand going through the top page, it's that there is no way to "break the seal" between the two pages unless you can curl up from the edge. I have to run to work, but I'll try to post a short video of this so you can see it in action.
User avatar
jck57
Posts: 376
Joined: 23 Nov 2009, 15:21

Re: Full Auto Brainstorming

Post by jck57 »

dpc wrote:
jck wrote: dpc: In my own experiments even moderate vacuum on a page tended to pull on the page underneath.
Interesting. Makes one wonder how that scanner I posted a link to is able to do it so well with vacuum? They claim that they can vary the vacuum based on the page detector sensor though. Maybe that helps? Thanks for giving it a try!
I have yet to try your curl + vacuum method. Previously, I experimented with a vacuum suction cup. Considering your idea, I thought since you're curling the paper no more than 90 degrees, you could have a flat section hit the paper for the vacuum port. I'll let you know how it goes.
dpc
Posts: 379
Joined: 01 Apr 2011, 18:05
Number of books owned: 0
Location: Issaquah, WA

Re: Full Auto Brainstorming

Post by dpc »

With the suction cup, are you actually getting vacuum through the page, or is the lift of multiple pages due to some other cause (i.e. static)?

The best would be a flat interface against the page for a proper vacuum seal. Something like a piece of square tubing with a slot or series of holes along one side would probably work well. This "vacuum bar" would lower and contact the page, then rotate counter-clockwise about the corner of the square tubing nearest the book's gutter. If the pages are stuck together, keeping that corner pressed down on the page stack while that rotation takes place will use the underlying pages' natural spring force to keep them from being lifted (or so the theory goes).
User avatar
daniel_reetz
Posts: 2812
Joined: 03 Jun 2009, 13:56
E-book readers owned: Used to have a PRS-500
Number of books owned: 600
Country: United States
Contact:

Re: Full Auto Brainstorming

Post by daniel_reetz »

I just received an email from Project Gado's Tom, who reminded me that they have an auto page picking type thing already open sourced. Might be worth looking into, at least for inspiration.

Edit: also they are apparently experimenting with variable suction, to accommodate different paper thicknesses.
Post Reply