[Xastir] Maps, again!

[Tom Russo]

On Sat, Oct 23, 2004 at 10:16:00AM -0700, we recorded a bogon-computron collision of the <archer at eskimo.com> flavor, containing:
> On Sat, 23 Oct 2004, John Ronan wrote:
> 
> > Also I had a though re the 'stitching' of the maps while on my return
> > journey last night.  Its really not necessary for me to do it (but I'd
> > still like to try).  As the event I'm trying to get organised for is an
> > event where we give support to the Hillwalking association. We are
> > planning on putting 2 trackers per walk, one per transport vehicle and
> > one one the Mountain Rescue Vehicle.  Basically the entire mountain
> > range will be on one scanned section, so all the 'action' will be in a
> > relatively small area of the map. So I don't really need to do any
> > stitching.
> 
> Actually, Xastir can do the stitching for you nicely.  Just crop all
> of your maps at an exact place, the grid lines should do nicely.
> Tell Xastir to load all of the maps you're interested in, and
> assuming that you've georeferenced each tile properly, Xastir will
> lay them side by side.  You can go as small or as large as you like
> with the rectangles.  Smaller rectangles might be easier to
> georeference and have less distortion across them, but then of
> course you have more work to do to georeference all of them.

Unfortunately, as far as I can tell, xastir with .geo files doesn't actually
do things quite right for anything other than lat/lon rasters --- it tries
to use two tie points, hardly enough for proper handling of the transformations.
If the scanned map is not in an unprojected coordinate system (and it probably
isn't), even if all the scanning errors were eliminated just slapping a 
.geo file around it won't really do the Right Thing, and you can see 
significant transformation errors, especially if you're using something 
like UTM maps well away from the central meridian of the zone.  

Even with a good scanner, you're bound to have some small error just from
the positioning of the paper on the scanner.  This can be corrected for with
enough effort.

I have had some luck using GRASS to georeference scanned maps for digitizing
vector layers.  I scan in the map, use the image rectification modules to 
get it right, and then use it as a background for the digitizer module.  
You could go that route and skip the vector creation part, just
sticking to the raster creation steps.  

GRASS is not easy to use, but the result is that once you do the labeling of 
tie points and rectify the image, you've got a GRASS layer that can be 
exported with proper georeferencing --- since they use gdal for output, one 
can even get a properly made geotiff.  It's a lot of work, though, as you have 
to first create a location with the right projection and datum to match your 
map, then read in the image into a *different* grass database just for that 
image X-Y coordinate system, associate points in the map with their real world 
coordinates in the target location, and keep iterating the ground control 
point selection until the rms error is low enough (this is the time consuming 
and frustrating part).  You need all the right information, too: DPI of the 
scan, projection and datum of the source map, and good tie points.  If your 
map has labeld grid lines for the coordinate system of its projection that are 
nicely visible in the image, that last bit might be less troublesome than the 
ones I've had to deal with.

-- 
Tom Russo    KM5VY     SAR502  DM64ux         http://www.swcp.com/~russo/
Tijeras, NM  QRPL#1592 K2#398  SOC#236 AHTB#1 http://www.qsl.net/~km5vy/
 "The truth will set you free, but first it will piss you off."