Phil, >From what I can see the technology is used to detect a limited set of signs only (e.g. Speed limits), rather than reading town names/road reference numbers/etc. There are a few different ways to do this which you can find by searching online, but generally the steps are:
1. Detect presence of a sign. 2. Determine shape of sign (circle, triangle, diamond, ...). 3. Determine content of sign (speed limit, no parking restriction, ...). 4. (Optionally) determine location of sign using GPS and position in images. A good demo to look at is http://www.youtube.com/watch?v=ywyOWLXR01c In this example they are using colour filters for step 1, and are therefore only picking out red, yellow and blue signs. Neural Networks are used for step 3. The journal article is available online, but it doesn't look like the software is. Rob p.s This could make a good GSoC project, but you would need someone who is knowledgeable about image processing (particularly colour and vectoring) and neural networks.
_______________________________________________ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk