New to the Archaeologist’s Tool Kit: The Drone
nytimes.com
nytimes.com
Would also be grateful for any papers or articles which discuss the algorithms involved -- especially anything with implementation tips/strategies. I understand the basics of photo stitching, but structure from motion is a bit more nebulous and something I've been meaning to learn more about.
On a sort-of related note. You might find this interesting. Record video from drone and overlay it on mapbox maps
For free (but not enitrely OS) you can look at VisualSfM[1], which is a GUI which pulls together a variety of separate tools to generate SfM. You can run all these tools seperately too, VisualSfM just vastly simplifies things. The author (Changchang Wu, now at Google) has a couple of interesting papers on the algorithims[2].
Both the commercial and the free software largely give you the same quality of result, the commercial stuff is just usually much much easier to work with.
As always, quality of your source imagery is paramount. I can't over emphasise how much geo-referenced images will help you. Second best is some "ground control points" (ideally with GPS locations). I've had limited success with completely unreferenced datasets but it is a slow process.
My workflow with GoPro video is to first defisheye the footage in CineForm Studio (there are other ways, but I find their defisheyeing to be pretty good), then I grab a frame every second or so from the footage depending on how fast I was flying around the object of interest. Too much data can almost be as bad as too little data!
[0] http://www.agisoft.ru/ [1] http://ccwu.me/vsfm/ [2] http://ccwu.me/vsfm/vsfm.pdf