I was previously involved with RMIT University’s remote-controlled blimp, and I’ve been to demonstrations of small UAVs, but I thought the entry cost was prohibitive. The Belgian Gatewing UAV, for example, has a $100K price tag, which includes the flight control and imagery stitching software. It uses an off-the-shelf consumer digital still camera.
Arko Lucieer at the University of Tasmania has been doing very exciting hyperspectral work with an octocopter mapping vegetation on Macquarie Island [the TerraLuma project, http://www.terraluma.net/ ]. Macquarie has a very unforgiving climate, very windy and wet, but he has been mapping over a large area to about 20cm precision.
I’ve been experimenting recently with a remote control quadcopter, the DJI Phantom, which is an off-the-shelf product. There is a remarkable amount of information about build-your-own quadcopters, of which I was previously completely unaware, so this was my first toe-dip, and one in which the learning curve has been steeper than I expected.
The DJI Phantom is less than $700, and is GPS-equipped, so it can hold position and also [fortunately for me] has a return-to-home function if it loses contact with the controller. It is a consumer product, but a very sophisticated one, unlike the AR Parrot. It is very easy to control, remotely-piloted, but not programmable [i.e. it is *not* a UAV]. To use it for gathering imagery, you have to pilot it yourself. Very soon you start adding things to the outside of the airframe; a camera, a video downlink transmitter and receiver so you can see what it is seeing, an on-screen-display module for telemetry, etc., etc. The payload limit is around 900 grams, but you quickly end up with a forest of wires, video and power, and instruments, as you gradually add pieces. I think it would be a great teaching tool, in the sense that you have to understand how the various pieces work together, rather than being a fully integrated product.
If you have a consumer-level camera with GPS tagging, it is easy to acquire imagery, and then use open-source or commercial stitching programs to create a mosaic. You have to quickly understand things like FOV, image intervals, flying height, but the positional stability of the quadcopter under GPS control is pretty good. If you wanted to, you could take a snap, move along, take another, etc. Getting a usable mosaic is non-trivial, because you can’t pre-program the flight path, but you do learn how to do point-to-point flying, and you also make sure to take a lot more images than you think you’ll need, because you *will* need them.