[talk-au] Accuracy of overhead imagery vs GPS traces

[Liz]

On Sun, 2 Dec 2007, Matt White wrote:
> What do we regard as more correct - the yahoo imagery or GPS traces?

I bugged my offspring for an answer and he provides the following

From Adrian Billiau, Surveying, University of South Australia,
Second year.

The most accurate the imagery can be is the same as the size of it's
pixels. Better accuracy then this can be determined for long linear
features such as roads, but it requires some very patient digitising.
In addition to the error imposed by the pixel size, there is another
error, that of attaching the image to the earths surface, or
georeferencing. Georeferencing uses a series of points on the earths
surface of known location, and these are defined in the image. A
mathematical model is then developed that closely fits the image to
these points. Closely is highly relative, but less then a pixel is the
standard. This means that a point in the image can be misplaced by up
to pixel size * georeferencing error, or 2 * pixel size.
        A GPS trace accurate to +-4m is then as good as imagery with a
pixel size of 2 metres. Example imagery of that standard is quickbird,
which is the source that Google uses for it's in town Google Maps
satellite overlay.
        My honest suggestion would be to take another trace along the
same road, and look at which of those 3 sets of data are the closest.
The true road position is likely to be the average of the 2 closest
traces. GPS has it's own regular errors of positioning, and it may be
that your GPS was suffering from one of those during the trip.
        For a direct answer to your question, I believe that a properly
functioning GPS unit will provide better accuracy then the imagery.