Aerial photography
From Photo-3d
The following was contributed by Bruce Gold:
Point camera perpendicular to direction of flight. During cruise, assume craft points where it’s going. Not so during maneuvers: take-off and landing are nose-high. Also not turns, aerobatic maneuvers. Probably not a problem. Just keep it simple.
Light aircraft usually cruise at 100-150 f/s 30-50 m/s. Guess the distance to your subject and figure the interoccular basis you want; 1/20 to 1/50 of the subject’s distance. Then decide the number of seconds between pictures. Try taking a quicker triple instead of a slow pair of pictures to give yourself more options later. Direct sunlight on aircraft window causes gross flare and reduction of contrast. Camera gets shady side of craft. A polarizing filter helps contrast, too. Forget long lenses; too much vibration.
Example 1: Mountain is 2 miles or 10,000 feet away. <--Remember you’re guessing. You are travelling 100 f/s. 10,000/50 = 200. 10,000/20 = 500. 200/100 = 2 seconds. 500/100 = 5 seconds. A pair should be exposed around 2-5 seconds apart.
Example 2: Forest nymph is 500 meters away. You are travelling 50 m/s. dist/ratio = 500/50 = 10 m interoccular distance or 500/25 = 20m. i.o.dist/V = 10/50 = .2 second or 20/50 = nearly 1/2 s. (Remember the precision of the original data before condemning this coarse arithmetic.) Image pair should be < 1/2 second apart. You are too close. You probably can’t take them that fast. She might throw a rock and knock you out of the sky. (Nymphs’re tough!) And you might crash because your pilot is obviously looking at the nymph instead of where you’re both going.
Clean the windows thoroughly, inside and out. Bring along cleaning towels anyway.
Barry Levin adds:
The quick answer is use the one-in-thirty rule: 6000/30 = 200 foot separation. 100 mph is 146 2/3 fps. 200 ft/146 2/ 3 fps = 1.36 seconds delay. So your shots with the longer delays may be difficult to fuse. The answer above is correct if you were pointing the camera straight down. If you were pointing it out an an angle of, say, 30 degrees from the vertical, then you would divide the 1.36 seconds by the cosine of 30 degrees to get 1.57 seconds; not enough difference to worry about. By the same reasoning, if you were 60 degrees off the vertical, then the time delay would be doubled.
John Roberts adds:
Those angles should apply to the *bottom* of the field of view, rather than the center. (Watch out for clouds - if they’re the closest objects, they should be the reference for calculations.) I would also recommend trying a range of interocular/distance to subject ratios of 1/30 to 1/50-70 (or even 1/100). I seem to get the best results for hyperstereo using a ratio of 1/50 to 1/70 (using an SLR camera with a 50mm lens). If you can afford the film and processing, it doesn’t hurt to hedge your bets by taking a series of photos, so you can pick out good stereo pairs later - after all, airplane tickets are pretty expensive compared to the cost of film. I sometimes take three or four pictures as rapidly as I can advance the film (with an autowinder, I’d be a little more scientific). If you find your stereo pair has too much depth, you can often salvage it by cutting off the bottom of the photo, which in a photo from an airplane has the closest objects.
Another very rough rule of thumb I sometimes use when I don’t have other criteria for judgement: the second shot should be taken when the view shows a barely perceptible difference from that of the first shot. It helps to remember this rule of thumb even when you calculate the spacing more precisely, since it’s intuitive and helps in avoiding order-of-magnitude errors.
