Hi, I see in the Wiki that you are planning to use Rhuus or a similar system to mark the sites. depending on the timeline of these activities, you may be able to use Mapknitter's annotation functions, as will have them re-written by end of summer. It would be great to have a test case community to work on the annotations interface with.

Wow! That sounds great! I'll try to add some more information to the wiki to give people a better idea of the general project parameters and what are current limitations are. Currently the biggest challenge we're having is the limited network coverage in the field sites. Even if the iPads the kids are using had 3G, it wouldn't do any good as there's no service anyway. When I first wrote the wiki we were looking into Rhuus for its cross platform/offline compatibility. We eventually realized that there are other options available now, like Mapbox, that weren't possible yet when Rhuus was first created. Do you think this is accurate?

The annotation feature on Mapknitter would be a great way to view the geotagged field data on its own layer. Especially if we're able to capture aerial imagery simultaneously with the field collections. Although, there will likely be days, as there was today, when samples are collected without companion aerial imagery. We would have to make sure to clearly identify orphaned data that may be overlayed on an earlier basemap.

Is it against the TOS to use a .kml from GE as a reference? If not then maybe I can try it out sooner rather than later?

One of the other features we're adding to mapknitter is that it will recognize the geo coordinates stored in images' EXIF data, and the ability to embed non-map images. This could work well for cross-platform offline use-- photos could be taken in the field with cellphones and brought back and uploaded into mapknitter, where they would automatically be placed. While the field notes wouldn't automatically end up in mapknitter, the images could then be annotated in the platform.

That sure would be amazing if all we had to do was take a picture with the iPads the students are using to get coordinates! But my guess is we would still need a GPS chip to get an EXIF tag and the Wifi-only iPads have no embedded GPS module as far as I know. We actually used an old iPod touch for our initial KAP experiments so I'll see if I can get accurate EXIF data out of it to see if it works. If not, I'm wondering what the next best geotagging method might be? Right now the students are using some old Garmin units to take a reading and enter the coordinates manually. I have all kinds of Arduino-compatible GPS modules handy and I figured I could rig up a few DIY geotaggers, but the (non-jailbreakable) iPads would pretty much rule out any connections outside of Bluetooth. Somehow, I don't see the equation of "tidal mud flats" + "serial port comms"+ "bluetooth pairing requests" as amounting to anything but frustration. That would be amazing if we had to take a picture of the iPad entry with an Android phone just to get the geocoordinates! Though I imagine it wouldn't help my reputation for being a Linux snob...

If you're wanting to use a multicoper for mapping, I'm working on automating many parts of that workflow for my dissertation. By the end of the summer I'm hoping to have the post-processing routines finished so that it is probably to fly a quadcopter manually and have it collect photos which are then merged and converted both into georeferenced 3d point clouds and georeferenced orthophotos. I could imagine that that's interesting for baseline mapping. My software will work with multispectral imagery from a single camera. Right now I'm waiting for drone parts to arrive in the mail in order to get going and have just been working on some small matters for the sensors, as documented in part here: http://erget.wordpress.com/2014/05/02/producing-3d-point-clouds-from-stereo-photos-tuning-the-block-matcher-for-best-results/

That's more for low-proximity mapping, but it's planned to expand that for aerial mapping as soon my parts arrive.

That sounds like a very interesting project! Have you submitted a research note for it? I for one would definitely be interested in exploring DIY quads and grassroots mapping in more detail.
There is certainly a wealth of online documentation on sites like diydrones.org etc... but I can't think of anything specific to diy mapping. I'd also like to look deeper into the long term cost benefits of balloon mapping (balloons, helium rentals etc...) vs the upfront cost of an open source quad build like MultiWii or Arducopter. I wonder if there are certain scenarios where an upfront cost of $300-$500 makes more sense than renting a helium tank every time. It would also be interesting to see if there are situations where forest cover/other obstacles prevent tethered flights like those of a balloon or kite but might be possible with a quadrotor. It would also open up more possibilities for real-time capture as "FPV" telemetry modules are getting more and more popular in the RC community. Such capabilities are much more challenging for unpowered flight as it would require a dedicated power source and would add significant weight. Concerning your goal for collecting georeferenced orthophotos. I wonder how difficult it would be to preprogram that functionality into a flight plotter like MultiWii? I could picture a format wherein one draws an outline of the area they want to map, and the program would plot its own path according to the focal length of the camera and other factors that affecting accuracy. You might not even need a gimbal in such a case as there would ideally only be one angle of view and that could be adjusted through the quads own gyro/accelerometer. I've never worked with Arducopter but MultiWii is Arduino-based and relatively easy to customize once you figure out the configuration files.

@code4maine-- certainly, doing bluetooth coupling with a home-built Gps unit and the rest sounds like a bit of a pain in the field. I'd just take two photos-- one of the thing you want, and a second of the GPS unit, and then manually place the images upon return. Its not ideal, but its dead simple.

@erget-- are you using entirely matched stereo pairs, or are you moving to do a SFM-style unmatched parallax extraction?

We have a big team working on completely re-writing Mapknitter this summer. EXIF-based image placement is a central thing we're working on, which should reduce the processor intensity of doing point-cloud generation. We see that as one of the next steps for mapknitter. Would you be interested in working on your software in a server-side implementation?

@code4maine a $300-400 quad will hardly fly long enough to map much (10 minutes), and is likely to get blown going to get blown away in the wind. I'd recommend going to a fixed-wing craft made out of expanded PP foam so that it can survive a few crashes, like the MAJA. None of these platforms are going to lift a ton of equipment though, their payloads are usually around 500g-1kg. I don't know much about multiWii, but Paparazzi is a really robust automation platform you should look at. And don't forget to get FAA clearance for educational use.

If you're on the coast of Maine, why not just use kites? dead simple, cheap, and they can lift many pounds of gear, no FAA clearance required.

$300 would admittedly be a challenge, but a $500 quad with decent lift and up to an hour of flight time is not only possible, I've seen it done a number of times. The only reason I haven't done it is I'm still aiming for that $300 mark! A lot of it comes down to the ESC/Motor combination and whether or not you can get away with one of the new all-in-one ESCs... (I won't clog the comments up with links but I've been following a couple interesting builds if anyones interested). The point about wind is well taken, and it actually leads into my answer to the suggestion that we go with a kite instead. In short, KAP is pretty much all we've been doing because of the logistical issues mentioned earlier about obtaining helium. I actually went into this very issue in detail in another research note from yesterday- http://publiclab.org/notes/code4maine/05-06-2014/simple-weather-alert-web-app-for-grassroots-mappers. The problem is, even living on the coast doesn't guarantee sufficient wind conditions all the time. For example, as of this writing I am right on the coast and windspeeds top out at 6MPH. Our experiments outlined in the other research note showed us that winds need to peak around 15mph for us to get the kite to a steady altitude. The thing that I like about the Kite/Balloon combo is that they compliment each others strengths/weaknesses. In other words if there's not enough wind for a kite, then it might be perfect conditions for a balloon. This would work out perfectly if we were in the habit of keeping a full tank of helium in the closet, but lacking this requires an element of planning (which was the motivation behind the other note). With quadcopter tech getting ever affordable and capable, there may well be a role for it to fill some of the gaps created by notoriously unpredictable New England weather and a worldwide helium shortage. In short... If it hasn't happened already, Moore's Law would eventually triumph over Newton's...

Of course, there is one form of law that nullify's Moore's law ... That is the very real law enforced by the F.A.A! I must admit that I was operating under the assumption that citizen science/educational use would be covered under "personal hobby activities" so long as its within line of site? Do you need to get clearance to use the images for mapping? Or because it is working within a school context? I certainly hope they get that straightened out by their deadline next year. Otherwise we'll just have to pretend we're making deliveries for Amazon...

You're right that if you put together a kit there are some quads around $500 that are fairly capable, but there is some significant setup involved usually.

15mph winds are pretty high. what kite are you using? one of the 9' Delta's in our kite kit will lift a small camera in 7mph winds. We don't sell a super-low wind kite. The Fled will definitely lift your equipment in 6mph winds.

Then there are Rokkakus, which are more "expert" kite with bridle adjustments and bowing adjustments needed. but then again, flying quads isn't all automatic. G-Kites 7.5' Rokakku, or SkyDog's 7.5' Rokkaku would do alright, both are fiberglass sparred and not as light as they could be. Now, if you REALLY need some pull, cobra kites sells Premier Kites 9x10' carbon fiber sparred Rokkaku that will very likely lift a LOT in very light wind. It is $300 though.

The FAA requires universities to get permits. I don't really know what the deal is for secondary schools. We, as a non-profit, aren't allowed to fly them.

@code4maine - I haven't submitted a research note, unfortunately. What you describe, however, is pretty similar to what I'm planning. I have to say that I'm very surprised about the possibility of putting together a multirotor aircraft that can stay in the air for an hour - that's way outside of anything I've heard of, although I work much more on the software than on the hardware side of things. For such long surveys, I would also think that a fixed-wing aircraft would be better suited.

@mathew - I'm definitely going to be using SfM for 3D reconstruction of aerial surveys. In the very long term, short-range 3D reconstruction from stereo images is something we're wanting to try out for close-range navigation - a way of doing some very rudimentary, zero a priori knowledge SLAM. It's actually planned for much later in my PhD, but since I don't have a drone I wanted something I could work on now, so I just built a stereo camera and got going with it.

I would love to contribute to Mapknitter. I think that would help a lot in simplifying the workflow (so the code actually gets used ;) ), and since Mapknitter can also be installed stand-alone it would still not require that the user have an internet connection, which is an important requirement for my project. Is there a development road map, project guidelines, etc., and how can I coordinate with you best? I see that the sources are on GitHub, but I think it would be good to speak about the details before I send any pull requests, etc.

If you have Garmin GPS units, there might be a better way for students to georeference photos. If the Garmin units save track logs, with a waypoint every 10 seconds or so, these can later be matched with the time stamps in the photos (gotta set the camera clock). Lots of programs will do this and write the Lat. Lon. to the EXIF header of all the jpegs. Notes and samples can also be hand labeled with precise time stamps, and these can be matched with the Garmin track log. Some cheap, old, tiny, digital voice recorders attach time stamps to recordings and have USB ports for offloading the digital audio files as wav.

So my standard field protocol for some data collection is to have the Garmin keeping a log, talk into the Olympus VN-6200PC, and when I stop recording, take a photo of the sample, site, or whatever. RoboGeo later writes Lat and Lon to the jpegs and attaches the audio recordings to the photos.

Of course the audio files are analog data and have to be transcribed, but that could probably be done with voice recognition software.

A phone app could do all of this, but how many students have phones these days. Wait, what?

cfastie- That seems like the best approach yet. I'm going to forward your suggestion to the teacher in charge of field collections though we might need to break it down a little more as he's not so technically inclined.

mathew- The kite we've been using is the HQ Power Sled, and has proven to be more capable than most models at low wind. I don't think there are any issues with the kite, though it occurs to me that there might be some serious shortcomings in my use of baseline windspeed metrics if 6mph is considered sufficient. Unless there really are kites out there that can lift off in any conditions no matter what, then as a matter of course there will be some days when there is not enough wind, even on the coast of Maine, to lift a kite in the air without mechanical assistance. My point about the UAV+Balloon+Kite comparison is not to judge once and for all, which one is best, but rather to show how each tool has its own unique capabilities that can be used to complement the others if used correctly. The goal is to match the capabilities to the field conditions that might demand it. I figured wind conditions would be a good measure as it certainly has an effect on whether one uses a kite or balloon. However, I'm beginning to think that basic wind measurements may not be a very accurate tool for determining which tool to use. This could be of major significance to my other project I mentioned earlier where I've been trying to set up a weather alert webapp for grassroots mappers.
I've decided to try another experiment which I'll document on my other research note- http://publiclab.org/notes/code4maine/05-06-2014/simple-weather-alert-web-app-for-grassroots-mappers

My usual method for determining if conditions are sufficient for kite mapping is to simply look out my window at the height of the waves. This is actually pretty reliable when you consider larger waves must be generated by relatively consistent winds. Like yesterday, there does not appear to be nearly enough wind to fly a kite today. Yet some weather sources still report windspeeds up to 7mph. So I'm going to try a couple different things... First I'll take a picture of the waves measured against something for reference. Then I'll take readings from a couple different windspeed apps and log the measurements. Then I'll sample weather reports from a few mainstream sources. Finally, I'll see if I can in fact get a sled kite in the air in present conditions (which would also be a good test for whether or not wifi-only iPod/iPads will stamp Exif data if nowhere near a wifi connection since I'll be using an old iPod touch).

Nice thinking Chris.

@code4maine I don't mean to discourage you from acquiring a quad, I just really like kites, and think of coastal areas (even without strong winds) as ideal kite spots. At 64" x 41", I'm not surprised that the HQ Power Sled doesn't lift your camera without a stiff breeze. Sled kites often need a good wind to keep their pull consistent, though i would expect 9- 12 mph would be sufficient given its surface area.

@erget awesome! can you join the plots-dev list and we can coordinate on there? A while group of Google Summer of Code students are joining our Dev team to work on rewriting mapknitter and we should loop you into our meetings.

Don't get me wrong. I didn't mean to come off as "the drone guy" (and every DIY tech forum seems to have one these days...) I think my intended point might've been obscured by my attention to UAV potential. I meant to say that every tool, be it kite, balloon, submarine, aircraft carrier etc...will have its superpower and its kryptonite. But these relative strengths and weaknesses strongly complement each other. So where windless days may be the kite's kryptonite, it would be the perfect place for the weather balloon to unleash its superpowers.
Far from advocating Quads over anything else, I'm not even sure if they have any purpose at all that can't already be accomplished more effectively and affordably through another tool! But this is what I wanted to examine. Perhaps quadrotors could fill an important niche mapping places that are hard to reach with a tethered balloon or kite? For the time being at least, its likely to remain in the hobbyist/enthusiast domain and as such would be highly unlikely to ever match the accessibility, affordability and ubiquity of the KAP techniques. However, it does make me wonder if there are ways to tap into the collective expertise of other hobby/enthusiast communities for specific roles. HAM radio comes to mind?

One last question... We'll be test flying either a balloon or a kite tomorrow afternoon at the research site... I've rigged up a Raspberry Pi w/NOIR Cam into an Otter Box enclosure akin to a rig you posted on a while back. Is this something you would recommend after doing it yourself? Are you still using the same Python timelapse script or have you found another way to remote trigger the function?