I'm working in Peru and considering lofting a camera with a balloon (http://breadpig.com/products/balloon-mapping-kit) over some 1-ha plots. Anyone ever do this before? Won't have as many pics to work with as with a drone, but I don't see why it wouldn't work. Anyone see any potential issues with this?
You might want to check out some of the work folks have done at Public Labs: http://publiclab.org/. They're the ones who make the balloon you linked to and there are some example of work in Peru. Balloons work very well for aerial imaging. Some drawbacks are that most of them use helium which is a little expensive and not always easy to find and they don't work well if it's windy. You also have to walk them where you want to go and that can be difficult if you're in a forested landscape. Some advantages are that are easy to use (although some practice is a good idea before you head to the field), they can stay aloft for a long time and they are excellent for short-term monitoring of a small area.They also tend not to be as heavily regulated as UAV platforms.
Excellent advice- but do note that balloons at altitude are also in the "gray area" of airspace...
I am not familiar with the rules in Peru but in the US as long as the balloon diameter is less than 6 feet and the gas capacity is less than 115 cubic feet there are no altitude limitations for tethered balloons.
Thanks Ned - just posted my project up there. http://publiclab.org/notes/allie_shenkin/07-13-2013/canopy-structur...
I just read the publiclabs project description, sounds like an interesting plan! I think you will need a lot of overlap to use Photoscan with balloon photos collected in the manner you describe. Photoscan likes to have lots of overlap (often > 80%) and usually does not like working with photos taken during times of ascent and descent when the baseline is parallel to the camera nadir view. In our early kite aerial photography work, link here, I marked out 100' lengths on the 1000' of kite line to get a rough estimate of at least how far out the line was. You should probably do something similar with your balloon line so that you can better estimate the camera field of view to plan for the right amount of overlap for Photoscan. We use a simple Excel spreadsheet to estimate field of view, resolution , etc. with several of our cameras for mission planning and I can share that if you are unfamiliar with such calculations.
How are you going to georeference the models? We have done a lot of work on that topic especially when working in continuous canopy forests in Eastern Maryland that are imaged with a copter system, see our recent open-access paper in Remote Sensing of Environment for details on our georeferencing approaches, link here. Some of that code is available on our code page.
Good luck! I look forward to hearing more about your progress.
One more thing! You might consider *somehow* using multiple balloon tether lines. For a few months I tried raising up 36" latex balloons up through the canopy to serve as reference markers for georeferencing. Even in *apparently* no wind days, as soon as the balloons broke above the canopy they would invariably blow sideways as they met the boundary layer conditions above the forest. This made using the balloons as targets basically impossible as they would blow around constantly during a copter photo acquisition.
Thanks for all your thoughts! They definitely got me thinking. The plots I'm planning to measure are relatively small (100x100m), so getting photos that overlap might not be all that hard. Along those lines, your 'multiple tether' comment lit an idea. We use bigshot slingshots down there to get ropes up trees for climbing. It might be possible in some cases to shoot an additional tether over the canopy and down to another point in the forest. That way we might actually be able to slowly move the balloon along an axis and get multiple photos that way. Will be a bit messy, but could work.
I'm not sure how important maintaining balloon location is. In fact, I would think that letting it blow about a bit would be a good thing, getting photos from some slightly different angles. That is, of course, as long as the motion isn't blurring the photos.
I'd love to see your excel sheets if you don't mind. I've done some back-of-the-envelope calculations, but I'm seeing a 12 deg FOV for my SD870 at 28mm focal length - seems awfully narrow.
As a start, I was thinking of georeferencing using big trees in both lidar and camera scenes. Your bucket/balloon method for control points sounds interesting. It would add more work, and while it'd be good to have them in order to test the method, I might skip it if using the lidar data as 'truth' will suffice.
One question - what do you mean when you say "baseline parallel to the camera nadir view"?
Here is the spreadsheet. It isn't the most well documented thing, we are still working on getting all our info online. Please let me know if anything is unclear.
Perhaps you are estimating the focal length equivalent? What kind of camera is it? For the little point-and-shoot cameras that we use the actual focal length is about 4.9mm with FOV of 60 degrees or so.
As for the baseline comment, I agree that what is said is very confusing, sorry! What I meant is that the algorithms seems to prefer it when the photos are offset laterally from one and other relative to what they are looking at. So it is better if the images are moving horizontally relative to the canopy rather than up and down in the same place only. Does that make more sense?
Thanks Jonathan - yes indeed, I was using focal length equivalents. I'm getting 64 deg x 50 deg FOV's now. Your spreadsheet is nice! And thanks for clearing up the baseline confusion - makes plenty of sense - i'll pick a height and stick with it at any particular point. I'm wondering - is photosynth able to handle different heights in different locations? It'll also be interesting to see how it does with different lighting conditions, as we'll probably be floating balloons during different times of day.
Now I'm wondering what the best arrangement of points over a square 100mx100m plot will be. If I float at 167m altitude, my images will span 200m vertically. If I float on the 4 corners of the plot, and perhaps also in the middle of the plot, I should cover most of the plot with each image, have plenty of overlap, and each pixel is a little over 8cm square. I imagine my lens will be more of a limitation than the pixels.
Seems like a pretty reasonable plan to me, especially if you are somehow able to draw the balloon across each edge, but that might be a long shot! Please feel free to blog about your progress.