Ok. So I have been asked to research character recognition, using a webcam to scan documents (probably at a fixed distance to avoid focal issues)
The documents will have 6 to 10 digit print numbers (basically a ticket system) that I would have to convert to int and compare with a database.
Now, I know flash can do this, albeit with a large amount of effort - assuming I have a very limited range of fonts, and only 0-9.
Is this asking too much of flash (and of course, my skills!)
In terms of processing power, would it make more sense to snapshot the document and upload to a server, and process the Character Recognition with, say, C, C# or one of the various open source OCR apps?
Anyone with any experience?
In terms of processing power, would it make more sense to snapshot the document
Definitely, I would try it in flash only because that is what I normally code, but it is easy to send a jpeg to the server and then let some serious processing do the work this will allow for more extension in the future rather than a limited flash version.
I think I'd look for some 3rd party server side software that can handle it. There must be something out there?
I wouldn't attempt this myself, but hey, that's up to you!
It would be fun to try and d t in Flash and Flash would be more than capable.... but....
What are you doing anyway? is this for a fixed location, kiosk etc? or for home users?
If yo need it to be reliable your best bet is to send over for server side processing but you will need to consider potential server load and inbound data issues if you're using images and expecting a lot of people to use it. Keep the processing client side if you can.
Well, the theory is to use a webcam to scan tickets for electronic bookings, at the door of the venue.
The user will receive a number/character string and arrive at the venue; have the ticket scanned and accepted/rejected.
There is a specific reason for not using a barcode scanner which I cannot explain here- it must be a number/character string.
I thought limiting it to numbers would be the best way as there are fewer characters to recognise; I might even reduce it to the most distinct characters eg, 1, l, T and 7 are probably too similar, depending on the font used.
Thus there would not be a huge load on a server, but speed is extremely important, because people do not like to be kept waiting in a queue.
I've been reading up on edge detection and the various methods for OCR, and I don't see why it cant be done with AS3, but its going to be a hard one.
Anyway, the project is still at research stages. I'll see what I can come up with.
AS3 should handle it fine.
Sounds to me like you might want to run the scanner as an air ap and have some local OCR service or app that does the processing for it. It would be a fun as3 project but not neccesary unless someone really wants to pay for it.