[GRASS-user] OCR

[John C. Tull]

You can also try to select by color in Gimp. I've done this in the past with varying degrees of success. Gimp allows you to adjust the threshold on those types of selects. GRASS's r.thin and r.to.vect might come in handy if you get any useful information from this technique. Obviously, it is dependent on the map and whether or not features that you are hoping to extract have unique, or nearly so, colors.

Good luck,
John

On May 5, 2011, at 5:18 AM, Rich Shepard wrote:

> On Thu, 5 May 2011, Markus Neteler wrote:
> 
>> You could try with this software:
>> http://www.gnu.org/software/ocrad/ocrad.html
> 
>  Nah, that won't work. OCR, as the name implies, recognizes text
> characters: letters and digits. I've used gocr/jocr for years with varying
> degrees of success. Unless the typeface is simple (monospaced, for example)
> the software cannot make any sense of it.
> 
>  A map is not made up of characters, so OCR is inappropriate.
> 
>  What might work is to open the scanned image in The GIMP, open a
> transparent layer on top of it, and trace the lines of interest. Then run
> the file (you can try this on the original scanned map, too) through
> ImageMagick's 'convert' program to produce a .pdf version. This converts the
> image from bit-mapped to vector. Heck, convert might alao produce a .svg
> output that one could clean and tweak with Inkscape.
> 
>  I have a large digitizing tablet with 4-button cursor I'm looking to sell
> because it's been years since I've needed to digitize a paper map. Won't
> help in this case because shipping this 2'x3' digitizer would be quite
> expensive.
> 
> Rich
> 
> _______________________________________________
> grass-user mailing list
> grass-user at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/grass-user