[OSM-talk] Obscure African dirt tracks in Google Maps - how does Google records them ? Automated tracing ?

[Jean-Marc Liotier]

Isometimes work on Senegal from Bing imagery,and I'll hopefully convince 
some Senegalese friends to help me with names and POI. I took a look at 
the competition's progress and I have been very impressed : whereas 
Openstreetmap could formerly claim better coverage in Africa, it is now 
lagging in volume behind Google - take a look at 
http://goo.gl/maps/mH3pK (equivalent area in OSM at 
http://osm.org/go/azSB9oX--). Google now features impressive mileage of 
dirt tracks and residential dirt streets in obscure backwaters.

Surely the Google survey cars are not roaming those places - or I would 
be astonished. Does anyone knows how they do it ?A couple of years ago, 
I had posted my suspicions about them using automated tracing techniques 
in general and street grid detection in particular: 
http://lists.openstreetmap.org/pipermail/talk/2010-January/046539.html - 
but since then it seems that their techniques have improved a lot. Does 
anyone have information about the technology they use and how it could 
benefit OSM ? Maybe they just have an army of tracers, but automation 
seems more like how Google solves problems.

Last time played with it, the results of the Bing road detect API did 
not look satisfactory to me- but there may have been progress since 
those early days. But wouldn't road detection perform better using 
imagery in different bands -multispectral or even hyperspectral imagery 
? The Bing satellite imagery render uses false color images fit for 
pleasant human use- but surely the rawcaptured data is richer. Are there 
any remote sensing specialist here to enlighten about that ?

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk/attachments/20121016/fb193493/attachment.html>