[Geowanking] Probe based mapping of road network

[Sylvan Ascent Inc.]

>>I would expect that GPS 
>>readings would have error offsets in specific directions depending on 
>>environment like a nearby building or terrain shape.
 
Has anyone actually made a study of this? I would tend to agree with this, but does anyone really know? Are there reliable ways to compensate? I've always thought that you could combine GPS feedback from lots of vehicles to produce an accurate real-time traffic map for better routing, with accurate street maps as envisioned here just as a nice bonus.
 
Roger Bedell
________________________________

From: geowanking-bounces at lists.burri.to on behalf of stephen white
Sent: Mon 11/27/2006 8:43 PM
To: geowanking at lists.burri.to
Subject: Re: [Geowanking] Probe based mapping of road network


On 28/11/2006, at 10:50 AM, christopher.wilson at daimlerchrysler.com 
wrote:
> Goal:  Create highly accurate and complete digital maps of the 
> transportation network suitable for safety of life applications 
> with accuracy commensurate with future GNSS systems (decimeters).  
> It seems to me that this can only be done through a statistical, 
> probe based, approach since imagery and 'mobile mapping' approaches 
> are error prone with low revisit rates.

As with my previous posts to this mailing list, this is just going to 
be my unsubstantiated opinion.

A statistical approach to GPS is assuming that the error margin is 
perfectly uniform around the actual location. I would expect that GPS 
readings would have error offsets in specific directions depending on 
environment like a nearby building or terrain shape. Aerial imagery 
is about the only thing that I would trust for this kind of accuracy 
as it has the human factor of being able to eyeball for error.

Collaborative feedback (aka the community) would be the statistical, 
probe based, approach to identify problems. You can still use error 
margins to indicate the trustworthiness of such data, and gradually 
add in extra information from merging several ways of collecting the 
same data. I would investigate vision systems and image recognition 
as an approach, as road markings are very easy for a computer to 
identify. Lines and black bits.

Use GPS tracks to locate roads, then computer vision to extract road 
information. GPS tracks contain additional information like turning 
lanes, but trying to extract too much information from the same 
source becomes a problem in filtering trends vs outliers. If you can 
correlate against other sources of the same data, then you can make 
more concrete deductions as well as being able to more easily verify 
the data on the spot.

By aerial imagery, I don't mean Google Earth. I mean very high 
resolution source data (used to make maps) so you can see all the way 
down to the gum spots on the pavement. This is obviously not as easy 
to collect as a bunch of GPS tracks, but you're going to find it a 
very hard sell to attach life-saving importance to something that 
politicians and the public can't see for themselves.

For that reason alone, you'll need to conclusively prove that your 
tracks are accurate to that degree, which can only be verified by 
plotting against the reality of the roads themselves.

Steve (the unknown guy without a famous website).

--
   steve at adam.com.au


_______________________________________________
Geowanking mailing list
Geowanking at lists.burri.to
http://lists.burri.to/mailman/listinfo/geowanking


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://geowanking.org/pipermail/geowanking_geowanking.org/attachments/20061130/4e7e4058/attachment-0003.html>