[GRASS-user] orthophoto georeferencing stops working when adding points

[Shane Litherland]

Hi all,

GRASS 6.4.1 on ubuntu 10.04 64 bit, tcltk GUI (because of other wxpython
probs previously reported)

Have been trying my hand at orthophoto georeferencing.

I started with a digital image that was taken with a digital aerial
camera. Has most of the camera settings required for the orthophoto
parameters asked for in the steps. As the image is direct from the
camera I didn't have a fudicial marks etc to do one of the steps but
that didn't seem to be a problem.

Seemed to be going OK, putting in parameters etc.

Got up to option 7 (as per what shows in a terminal after i.ortho.photo
is run from the GUI menu

I have put in 4 control points using the monitor. For each one, i used
zoom-box to get up closer before positioning point, and using keyboard
to enter GCP co-ords.

I then did a fifth point. it seemed to work OK, but then when trying to
do zoom for another point, get messages:
WARNING: Unable to find <raster_i_am_georeferencing> (click mouse to
continue)
[so i click mouse, then second message]
WARNING: Unable to open raster map <ditto> (click mouse to continue)
[click mouse again, options along bottom of screen come back, so I quit]

I have tried repeatedly with fresh sessions of GRASS, to add more points
without luck. Also noted that only four GCP's are saved. each time I
start it up, I can apparently add GCP #5, and then get warnings after
that, but that #5 point isn't being saved to the control points file.

I tried a workaround: open up the GCP monitor/terminal, zoom and click a
point, then go and open up the control-points text file and manually
enter in that point co-ords and matching control co-ords. Go back to the
monitor/terminal and hit 'enter' to blank out the point I'd just clicked
on. Then tried zooming on another spot of the map, still got the errors.

Opened up a fresh monitor/terminal, and noted that the point I'd
manually added to text file was displayed correctly.

Did a zoom box, actually missed where I wanted so then used zoom-point
option to try and reposition the zoomed view, got the warning messages
again.

Dunno if it is part of same thing, but notice when I hit 'analyse' that
the errors for east and north are up around 657900 and -6603200
respectively, using UTM 56J wgs84 (i.e. metres). target points are
around 476900 and 7096500.
It's a while since I've done some georeferencing, but these figures just
seem too wrong. or does the error get calculated at the next stage and
these are really just dummy/filler numbers?

I tried another work around - run the 'georectify' tool from GUI menu,
as far as the i.points part I think? i.e. enough to put one example
point into the points text file. Then I copied the points from 'control
points' text file for ortho-photo process, over to the points file for
the regular process, noting the spaces, elevation etc that wasn't
relevant for the points file.

I could open that up in the georectify monitor, and the error shown for
each of the points was more along the lines of 0.7 - 26 forward error
and 5 - 125 backwards error (1st order rectification). That's hugely
different from what I reported above in ortho-photo analyse.

What have I done, that it was happy to input 4 points and after that I
can't work it?

If I have to manually add points to the text file, one at a time and
restart i.ortho.photo each time, I will eventually achieve what I want.
but that's time consuming, and absolutely not what the program was
designed for, right? :-\ On the other hand, I could go back to what I've
done before with the 'basic' georectify (i.points etc) but that won't
let me add elevation data to an image like orthophoto does??

Regards,
Shane.