I've been playing around with PanoTools and Hugin for two or three
years; now I finally want to use it for some serious quantitative
I'm interested in archaeoastronomy which includes the study of
purported ancient astronomical sites and would like to use
panoramic photos to define the horizon profiles surrounding some of
these sites. In this work it's essential that the panorama have good
geometric control in both azimuth and altitude. Given the well-known
tendency of Hugin to create solutions where the horizon is anything
but horizontal, I need some way to introduce strong geodetic control
into the solution. Arbitrarily selected horizontal or vertical
features in the images would not be enough for this purpose.
I have some fairly kludgy ideas for a solution that involves using a
partial equirectangular panoramic grid as an anchor photograph, to
which well-defined points in the others could be referred. The
anchor grid would be calibrated so that each pixel would correspond
to a specific angle. If each pixel were 2 arc minutes, then the
entire grid would be 360 x 30 = 10,800 pixels wide and with a height
corresponding to the height of the panorama. For naked-eye astronomy
an accuracy of 0.1̊ is about as good as it gets, so 2' pixels would
be more than adequate.
The idea would be to use a theodolite to measure the altitude of
control points that are readily visible in the photographs and are
well distributed around the horizon, and to compute the azimuth of
distant landmarks (using GPS coordinates of the camera position and
the distant landmarks). These altitude and azimuth measurements
would then be converted to pixel coordinates on the anchor panoramic
grid, and the corresponding pixels of the control points would be
identified on the photos in which they appear. For azimuth control
points, the match would be defined as a vertical line (since all
points with the same azimuth lie on the same vertical line); for
vertical control points, the match would be defined as a horizontal
line (since all points at the same altitude lie on the same
In stitching together the images I'd maintain the scale of the
original anchor grid, and I could then use Photoshop to read precise
altitudes and azimuths from the resulting panorama to within a pixel
for any points of interest (such as those on the horizon profile).
It would then be a straightforward task to convert pixel coordinates
back to altitude and azimuth.
I guess I have three questions.
First, am I reinventing the wheel and has someone already done this?
If so, I'd appreciate a pointer to an appropriate source.
Second, do any of you have any comments on my proposed procedure?
Are there any little flaws in my proposal? Is this the way to go or
does anyone have other ideas?
Third, is something like this worth adding as a feature to Hugin, so
you could mark points in the images and then directly enter their
altitude and/or azimuth to be considered as part of the optimized
solution, or is this too specialized a problem to be of general
Thanks much, Steve McCluskey