That's basically what I'm looking for. Here's why.

A colleague took a bunch of video out of a helicopter with the camera looking straight down. I'm trying to estimate the maximum off-angle measurement error due to roll and pitch. Since we have altitude, GPS speed, and camera frame-rate, we can make some calculations of hurricane impact on forested areas. But I don't have an error term for random buffeting (no gyro input).

So instead of worrying about instantaneous (frame-level) error, I'm thinking that I can just give a max error (altitude * (+/-) tan(off-nadir)) and then provide the mean estimate and confidence intervals. As a mapper, I hate it but as a ecologist, I'm okay with it. I've done a brief lit search and almost all of it is related to minimizing photogrammetric error. But that level of accuracy just isn't necessary, given the variability of the measured data.

See, tree measures (search term: forest mensuration) assume all trees are either cylindrical or conical (more accurate). But anybody who has tried to shoot a squirrel or dry a plank of green lumber knows that timber is funky-shaped and a cylinder or cone is just an easy approximation. When you consider the number and variability of species and stems along a 30 minute flight path, decimal places are just not important anymore.

If helicopters are generally within 5 degrees or so of that estimate at 500 ft, and whatever normal cruise speed is, I'm okay with it.

B