NavList:
A Community Devoted to the Preservation and Practice of Celestial Navigation and Other Methods of Traditional Wayfinding
Re: NG's "Midnight Fun"
From: Marcel Tschudin
Date: 2010 Jun 15, 11:49 +0300
From: Marcel Tschudin
Date: 2010 Jun 15, 11:49 +0300
George, you wrote: > This question takes on a bit of importance because postings appear on this > list, quite often, from proponents of the use of cameras for making > celestial measurements. The geometrical distortions discussed here, that > arise from portraying a spherical surface on to a plane array, add serious > complications to interpreting measurements, and deriving scale factors; > complications which are often neglected. It would indeed be possible to > calibrate the transfer function of a lens system and remap images on a > corrected grid, though this would call for a lot of work if a zoom lens is > being used. The "lot of work" may really depend on what you aim at. Calibrating a camera-lens-system in order to make them useful for measuring the altitude of the sun or moon above the apparent horizon to CN accuracies requires a series of photos along the axis of interest. The "lot of work" consists actually in measuring the dimensions in pixels, which - for statistical reasons - are preferably measured several times, e.g. about 6 times. By limiting ourselves in using for the measurements always about the same (central) axis in the photo and aim at having the middle of the distance sun-horizon in the centre of it, we ended up for one lens with a simple linear relationship and for the other with a second order polynomial for directly converting the distance in pixels to the altitude in moa; the scatter of the data points had in both cases a standard deviation of only about 0.2 moa. Using for the one lens instead of the second order polynomial also a linear relation ship increased the standard deviation to about 0.5 moa. These calibration photos were all taken at about the same time, thus at about the same atmospheric conditions. Considering that under different conditions this value may eventually be somewhat higher and considering further also all the other observational errors one may finally expect an accuracy of about 1 moa or even better, provided the photos are made the same way as the calibration photos and that the photos are measured using the same techniques as used for measuring the calibration photos. It becomes a bit more complicated if the axis sun-horizon doesn't have to be in the middle (but is still parallel to the middle line of the photo) and doesn't have to be centred. In this case the altitude is calculated by integrating the differential scale over the pixel range. The differential scales were in this case approximated with second order polynomials using for one coordinate ax^2+bx+c. The dependency from the other coordinate was approximated with a, b and c being also second order polynomials, like uy^2+vy+w. The resulting standard deviation (valid for the full x/y-plane) was about 0.5 moa. This analysis can be done in a spreadsheet by using e.g. a template containing all of the calibration parameters. The calibration itself results in more work since it means measuring the pixel-dimensions for several axis. When you wrote a "lot of work" you may have thought to calibrate for being able to measure any (oblique) distance within the x/y-plane. Would this really be necessary for CN applications? I could however imagine that there exist somewhere professional programs which would help to do also this sort of calibration and analysis of the measurements. Marcel