multi zone ocr

Discussions about machine vision support in GdPicture.
Post Reply
frenky600
Posts: 6
Joined: Tue Aug 11, 2009 7:16 pm

multi zone ocr

Post by frenky600 » Sat Aug 22, 2009 10:58 pm

Hi, i'm going to create a new class that retrieve some part of document with ocr function.
I try to use ocr plugin with rectangle coordinates with gdviever (left mouse button like your example) and then setroi command to capture text inside the rectangle without problem.
but, is possible to create a while loop to capture many zone in the same document ?
i would like to capture many zone without using GdViewer1.GetRectCoordinatesOnDocument function directly.
i would like to create a document template with all coordinates stored in database using two gdviewer, one to get rectangle coordinate and second to render many rectangle for all zone that i need to read (i use the second gdviewer to delete or add new zone)
When i tried to make this procedure, ocr fail to retrieve text in the right coordinates. example: if i get coordinates directly from GdViewer1.GetRectCoordinatesOnDocument and make ocr with this coordinates the text is right but if i store coordinates and reload in gdpicture the last image, the ocr get caracter in bad position.

sorry for my bad english.

User avatar
Loïc
Site Admin
Posts: 5881
Joined: Tue Oct 17, 2006 10:48 pm
Location: France
Contact:

Re: multi zone ocr

Post by Loïc » Mon Aug 24, 2009 11:02 am

Hi,

Are you dealing with PDF or images ?

Are you using GdPicture ActiveX or GdPicture.NET ?

Kind regards,

Loïc

frenky600
Posts: 6
Joined: Tue Aug 11, 2009 7:16 pm

Re: multi zone ocr

Post by frenky600 » Mon Aug 24, 2009 11:36 pm

Hi Loic, i'm using gdpicture.net.
I capture images from dr2580 (all series) canon scanner and render in tiff for best post processing command with your library.

yesterday i made many test with multi zone ocr and, probably, i found my problem source:

i scan the template image and then draw some rectangle (and store coordinates). when i apply the template to other scanned image (with the same layout) like order confirmation, invoice etc. the image can be shifted up, low, right or left due to scanner pickup registry or due to paper guide not correct set; so the ocr can't 'see' the right image portion and fail.
i think that i need to 2 ancor point to identify image shifting, to prevent that the coordinate stored from template image, is not in the same place of current scanned image.
have you some idea how to compare a little portion of scanned document and tempate document to calculate coordinates adjust ?

(do you think i'm crazy? :oops: )

your library is the best that i found on the web.....price/quality is very hight. i plan to buy gdpicture.net + ocr in seven day.
last test is with a multifunction photocopier (konica/minolta) that is very hard to control by twain standard. this machine was born only to scan to ftp but your library is very stable and i'm very confident.

best regards

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest