[OSM-talk] Imagery parallax error in high altitude areas

[Arun Ganesh]

It recently struck me while identifying mountain peaks in the himalayas
that something may not be right. All of us have noticed that the top of
skyscrapers is off from the base of the building owing to parallax error of
the satellite capturing the image at an angle. The average seems to be
around a 0.2m displacement for every 1m increase in height (based on
calculations made in a couple of cities in India). For an imagery tile
which has 1000m variation in elevation, various objects could be displaced
by as much as 200m from its real position.

This means that tracing mountain roads, streams and peak from imagery would
inherit significant parallax error displacement and there is no easy way of
accurately offsetting the imagery before tracing either. Has anyone done
any analysis on how bad such errors can be?

PS: I thought i'd look at the case of Mt Everest, which not surprisingly
has been marked using the high res bing imagery. I compared it with the
coords on wikipedia and there seems to be a difference of around 50m [1]. I
was expecting more displacement, but then again I dont know how accurate
the wikipedia coords are, If its based on google imagery, its got some
parallax error again.

[1]
http://www.openstreetmap.org/index.html?mlat=27.988056&mlon=86.925278&zoom=12&layers=B000FTF

-- 
j.mp/ArunGanesh
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk/attachments/20120401/178e7429/attachment.html>