h a l f b a k e r yBite me.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
(Grep: Unix command line tool that searches for words in text.)
A handheld unit containing:
(1) A camera with an OCR system.
(2) A microphone with a natural language parser.
(3) A laser pointer, directed by a small mirror.
(4) A gyroscopic stabilizer. (Or an optical one, as egnor
suggests.)
You hold the unit in front of a big and confusing map
(I got the idea while looking at one of Lumeta's Internet Maps), say a word, and the unit points to the word with its laser pointer.
I don't think we can OCR that well yet; but once we
can, since the number of ocr'ed words limits the vocabulary the natural language parser must recognize, this looks *almost* doable.
(?) Optical Correlator
http://www.opticalc...ical_correlator.htm
[hippo, Dec 02 2000]
Fourier goggles
http://www.halfbake...a/Fourier_20goggles
[hippo, Dec 02 2000]
Library of Congress Search
http://catalog.loc.gov/
In some way, shape or form, all 121 million items are on net [thumbwax, Mar 27 2002]
Optical Transforms of the Alphabet
http://scitation.ai...type=cvips&gifs=yes
A place to start... [csea, Aug 06 2009]
Sikuli
http://sikuli.org/
Another place to start... [not_only_but_also, Aug 01 2010]
[link]
|
| |
Yes, yes. Adapt it to scan microfiche and microfilm. |
|
| |
Neat. Maybe use a small keyboard instead or as well to get around possible problems with language recognition like multiple-language and dialect support. |
|
| |
Hook it up to a mechanical page-turner to do book searches (tiny portable unit docks to huge, ugly home unit?), provide other output, support advanced searches (even regular expressions, if you like), and make it available in a waterproof teal version. |
|
| |
Wait, I think they call what I described a 'scanner' and a 'PC'. Maybe the only new thing above is a teal page-turner.| — | Monkfish,
Dec 02 2000, last modified Dec 03 2000 |
|
|
| |
The camera would need to have very good resolution, or else have the ability to quickly pan and zoom around the entire map. You'd probably want to re-use the mirror you use to steer the laser. |
|
| |
You can probably ditch the gyroscopic stabilizer and use optical stabilization. Just watch the image, and keep the laser pointing at the word. |
|
| |
I think voice recognition is going to be the problem, not OCR. As far as I can tell, OCR basically works but voice recognition basically doesn't. (So use a keyboard or scribble pad or something, as Monkfish suggests.) |
|
| |
Possible problem: what if the map or whatever contains multiple instances of the searched-for word? I'm sure there's a way to deal with this more intellegently than to pulse your laser pointer back and forth rapidly between all occuances. Ideas? |
|
| |
<<added later>>
Hippo mentioned below using the fourier transforms of an image. Conceptually, what is that? I know about a fourier transform of a waveform, but I am confused as to how it can be extended to 2 or 3 dimensions.
I ask simply for my own elucidation. |
|
| |
I think bouncing back and forth between all occurrences is just fine. |
|
| |
If it bounces fast enough, you will see multiple spots, which seems perfectly reasonable to me. Or it could have a "next" button. |
|
| |
Optical Correlation is fascinating (and what I did for my final year degree project). Basically you correlate the Fourier Transform of the target with the Fourier Transform of the image - this allows you to match multiple occurences of the target at any point or scale in the image. See link. |
|
| |
If this was installed at the library, you provide the page turner and the search instruction. |
|
| |
I forsee problems with pronunciation. If you're looking at a foreign map, how do you know how the names of places are pronounced? In much of eastern europe they seem to have left out most of the vowels and in parts of Africa there are far too many. |
|
| |
Why bother with the laser pointer? |
|
| |
While I admit the coolness factor is off the scale, a
product like this might be more feasible in the short term
as a snap-on attachment to a Palm device: in essence a
snapon camera. You graffiti some text, the camera scans
the image and the LCD displays the relative whereabouts. |
|
| |
I thought this meant a grep-type tool for searching images and video. |
|
| |
ogrep --find-in video.ogg --find "Apple" |
|
| |
above line would search video.ogg for frames, segments, audio with the word (text or spoken) "Apple" or pictures of an apple. |
|
| |
...just needed to add a bun to this for the visual effect...+ ;) |
|
| |
Can we OCR well enough yet? |
|
| |
great to detect speed limits on roads and warn drivers. |
|
| |
Sounds like potential for an iPhone/iPod app. Sans the laser pointer, of course (would need to provide some kind of zoomable highlighted view on screen). |
|
| |
Another possibility is to integrate this into an [insert your preferred fantastic large-screen type phone with camera on the other side of screen] app. Search for the word by analyzing the camera image and highlight the word on the screen. |
|
| |