Auto Detect Roads – Bing Maps

As Steve Coast (OSM guru) moved to Microsoft a few months back more collaboration between Bing and OSM was to be expected. Word* came my way of an application developed by Steve Coast which allows a user to click 2 points on a road on a Bing Map satellite image. The tool then automatically follows the road and creates an OpenStreetMap XML file of the route. You could then upload your XML to the OpenStreetMap database I guess.

I experimented a bit with this and results were varied! As the autodetect algorithm relies on colour to decide where to place the next road node there are problems when the colour of the road in the image is the same as the surroundings… in the image below.

I did a bit of computer vision in my MSc (on image recognition of street furniture!) and I came across similar problems as these. Image recognition techniques based on colour are always more error-prone. But it’s not all bad. The example above isn’t really fair. To the human eye it is difficult to see the road between the hedges and trees in the centre of the aerial image. I ran the application on a wider road in the countryside and the results were pretty impressive. Also in semi-urban environments the results were good.

This application is specialist but should save OSMers time in collecting and digitising vector road data. As for attribution….well that’s another story!

*Thanks to Nick for this information.


This entry was posted in Uncategorized by Chris. Bookmark the permalink.

Leave a Reply

Your email address will not be published. Required fields are marked *

5 − = zero

Spam protection by WP Captcha-Free