Hacker News new | comments | show | ask | jobs | submit login
Drone Imagery for OpenStreetMap (mapbox.com)
171 points by bsudekum on Nov 19, 2013 | hide | past | web | favorite | 48 comments

This is insanely cool, and a bit crazy too. 45 minutes to map 10km^2, so two of them, swapping out batteries, you're talking one person mapping 90 km^2 a day. So a city like San Francisco (121 km^2) in a couple of days? Or San Jose (467 km^2) in a week? That is pretty interesting. Make your own map with license to your own assets pretty reasonably I'd guess.

The laws would have to change to allow you to fly these over populated areas. There's a reason the blog post is about flying over someone's private farmland. As a commercial company you can't even charge a farmer to fly over his land, the owner of the land has to do it himself.

The need for it isn't quite as much in populated areas either, at least for OSM's purposes. Bing's aerial imagery is fairly high-resolution and timely over cities (at least cities in the western world), and they allow use of it as a layer to trace OSM features from: http://wiki.openstreetmap.org/wiki/Bing

Push your local government to open up their data!

Eg North Rhine-Westphalia has published an "NRW-Atlas" which can be used to draw maps from it (no automatic imports, though), the data is really good in general. http://www.bezreg-koeln.nrw.de/brk_internet/organisation/abt... http://forum.openstreetmap.org/viewtopic.php?id=22926 (german)

hi-res Bing not always available even in Europe (e. g. http://tiny.cc/p27t6w) and Siberia almost not covered.

Attach transient, disposable cameras with precise positioning to wide-ranging self-piloting lifeforms not subject to FAA approval, such as city pigeons.

While you may be kidding, such a thing with a drive by upload to a local wifi network might be possible (no opinion on practical) but basically a camera, gps module, altimiter, and wifi link. Put feeding stations on Starbucks or near other free WiFi hotspots that dispense seed if the leg unit uploads a picture.

Here's an interesting idea: drones as really autonomous entities. A script with a bitcoin account sells aerial imagery to mapping companies and uses the proceeds to pay for its own hosting and for the upkeep of its drone. It's a company that really lives "in the cloud(s)".

In our present legal environment I guess the creator would still be charged with breaking the FAA regulations, but perhaps we'll live to see the day when our children will be held responsible for their own actions :)

they did that in ww1 :)

Easy solution. Don't charge the farmer to photograph his land. Instead charge to rent the drones to the farmer so he can "fly them himself" with supervision. Just code up an interface that gives him a big green "Go" button that sets off the actual flyover sequence.

45 minutes to capture the imagery. Also, 100 acres is 0.4 km^2.

In the U.S., the big advantage over government imagery is going to be timeliness. I'm sure some people also have uses for the resolution.

I was going by this:

"The eBee has a flight time of up to 45 minutes allowing to cover areas of up to 10sqkm in a single flight." -- http://www.sensefly.com/drones/ebee.html

I suppose it depends on the resolution required. Lower resolution allows you to fly higher, cover more area per picture and thus cover larger areas faster.

This is very cool. TileMill (https://github.com/mapbox/tilemill) is also one of the most amazing open source web apps I've seen for a while.

Completely awesome. I was curious so I looked up the eBee price. $12,000 - too steep for a hobbyist, but I'm sure that price will come down over time as drones become more widespread.

Not sure what you get for $12k, but know that it is possible to build an autonomous airplane, and a gopro for much less than that. Here's a friend's video for evidence:


His cost is order $2000.

You can do it for less than that, even. A Naza M flight controller costs $200 (http://www.uavproducts.com/product.php?id_product=21), and that's one of the more expensive controllers out there. The remainder of the components for a flying wing would run you $80-$120. You could replace the Naza M with a multiwii controller and write your own flight control program for much less. These are multicopter components but there are analogous components for gliders and planes. Then there's the GoPro, a used one of which might cost $150. I guess you can do this for an all-in price of less than $600.

$600.00 is a pretty solid estimate for a DIY, autonomous, man-packable flying wing with relatively robust performance.

I've done it for around $1100, excluding the cost of a laptop to run some open source flight control software, but ~$600 of that was a 9-channel RX/TX controller setup. Overkill unless you're doing some really crazy stuff.

It's not all sunshine and rainbows, though. I've worked closely with scientists that have had problems with SenseFly units in the field, despite the high level of polish in the product. I've flown similar sized UAVs to the eBee (all self constructed) in great environmental conditions (low altitude ASL, favorable wind conditions, etc.) and had great success. However, get these things in less hospitable environments, and things go sideways real fast. So, while Joe Everyman is getting closer and closer to just throwing a small UAV in the air and having it work, we're still quite a ways from guaranteed success out of these things.

Observation: I've never had good luck using a GoPro for mapping tasks. My suspicion was the rolling shutter, but perhaps I was just doing it wrong.

> Overkill unless you're doing some really crazy stuff.

Especially considering you can buy a 9XR + FrSky module for like $130 nowadays. This wasn't always the case though -- Rx/Tx solutions have become a lot less expensive since 2008 or so, as I understand.

Yep - I've got some footage from ~$130 worth of gear here:


I'd agree that you'd easily be able to replicate 95% of what they've done for under $600 (so long as you don't count the time you'd spend researching/building/learning - FWIW, I wouldn't sell one of these hypothetical $600 autonomous camera planes to someone else for less than a few grand - my time, warranty, and liability would easily add up to that.)

Honestly I would strongly recommend you avoid flying these without umbrella coverage. Odds of something bad happening are pretty low, but the financial costs could be catastrophic if it did. Read your umbrella policy to make sure it covers this type of loss as well (mine doesn't cover aircraft liability, but it defines aircraft in such a way that anything too small to carry a passenger is excluded, and thus covered).

It's definitely worth considering – but for me with that toy, I've considered and chosen to accept and ignore any theoretical "catastrophic" consequences, much the same as I ignore the risks of dying in a car accident on the way to the park to fly it.

The plane in that video weighs 128g (~4.5oz) thats ~20% lighter than a baseball (or cricket ball) and ~30% lighter than a frisbee. Sure, it's got a powered spinning propeller on the front, but it's pretty low powered (sub 6A draw at 7.2V - 45W tops) and more importantly I've got _much_ more control of it than people throwing balls or frisbees have - at least once they've left their hands.

I'm happy enough to accept the personal risk of flying this "toy" with the same sort of common sense rules that I'd use if chucking a ball or frisbee around with friends. Don't be reckless, don't "buzz" people, try to be obviously taking safety into account. Yeah, there's some risk - but a little bit of self control can easily reduce that risk to near-enough-to-zero (and I'm prepared if required to throw myself on the mercy of a court and argue that I'd considered and taken prudent precautions to avoid an accident in the case of a truly tragic incident.)

A bigger/heavier/more-powerful plane like the one in the OP would affect my decisions about when/where were sensible places to fly - but keep in mind the eBee in the article only weight "less than 700g" - that's less than 50% more than a football. Personally, I suspect there'd be a lot of times/places where other people occasionally kick footballs around that I'd happily fly a one of them - not without a little extra caution, but given suitable circumstances I'd fly a ~700g eBee on the same cricket ground as that video of my Mini Swift.

I got a quote back from a reseller for $25,490...where is your price from?

> $12,000 - too steep for a hobbyist

market segmentation. They used the "industrial" keyword. So it could have been anything between $300 for Parrot and $35K (airframe only, the whole system is $200K+) for Raven.

You certainly can do this for much cheaper. [0]

[0] http://diydrones.com/profiles/blog/list?tag=mapping

Man that's expensive. You could probably build something similar on a quadrocopter base for a tenth of that. Not a lot to the thing when you look at it.

A multi-rotor will ccurrently only give you 10-15 mins flight time with a payload. There are cheaper options (than the OP) that are also more practical.

It comes with a ready to go autopilot and, I think, the software to turn the imagery into useful data (the stitching and georeferencing).

Folks have been doing this for years, and for cheaper, with balloons and kites. Check out the awesome work at publiclab.org, where they've mapped the gulf oil spill in 2010, pollution at the Gowanus canal, Occupy encampments. Stitch the photos with the OSS Ruby tools at mapknitter.org, publish them with a public domain license, and they'll actually be added to the google base map!

Buy a kit for < $100: http://store.publiclab.org/products/balloon-mapping-kit

If only we didn't have to rely on nonfree software for the processing.

I used to work for a mineral exploration company that purchased an Ebee in the spring, and used it heavily this summer in Northern Canada (mountainous, extremely remote terrain). It's a pretty neat unit and only required about 1-2 hours of reading the manual and making a quick flight plan before we were ready to fly it. The software that ships with it is super duper easy to figure out. Though it may be too simple as some of the control we would have liked wasn't there.

That said, the gov't paperwork before we were able to fly it was a nightmare, but operating it in remote, helicopter access only areas (no humans/buildings/etc around for dozens and dozens of kilometres) did ease the requirements a tiny bit.

Generating DSMs from the geo-ref'd images within an hour or two of flying a flight plan was pretty awesome though. Being DSM (Rather than DTM/DEM) is kind of annoying with tree-cover, but in less treed areas is still better than the Government of Canada 30m DEM.

volunteers with army of drones can can cover cities, there can be a kickstarter for each city funded by its resident. I will fund for chennai and bangalore for sure

Prepare to have your expensive drone captured or destroyed by your government, and to be investigated as a criminal.

This is so cool.

I'd really like to do a drone services startup focusing on municipal govt's and agriculture. It seems like you could solve a lot of surveying/permitting issues for cities. With ag, special cameras can give farms a lot of data (I know this is being done already). I think becoming sort of the "Waste Management for Drones" could make a lot of money.

I was really bummed when I read that one of these costs about 12k. But then I realized how much money you could make taking this around the country, mapping out cities and licensing the images out. Heck, just taking private gigs for colleges, estates, parks, and things of that sort could make you good money. I think i found a new side business.

Check your local laws first.

I see lots of $5k+ multicopter/camera setups for sale at _big_ losses here in Australia, when people obviously got all excited about starting their own aerial photography business – then bumped into the legal reality of flying these things commercially as opposed to being a hobbyist.

If you want to accept money for flying RC aircraft here, you almost need the same qualifications as a commercial pilot _and_ a commercial airframe mechanic. There are a few dozen companies who've got the required qualifications, and they're _very_ protective of their turf. You _will_ get "noticed", and the authorities (CASA here in .au) will come knocking.

Wow, as a big multicopter and fixed wing enthusiast/hobbyist, I find this to be super awesome. It also literally hit home for me because I noticed that it was captured only a few miles away from where I grew up, haha. These technologies look very promising for things like mapping agriculture and providing recon in environmental disasters.

I live abroad and was considering pitching this as a low-cost alternative to satellite imagery and crowd-sourcing with the the government GIS teams and national computer research center ehre. Very, very cool to see it in action!

Cool, but you can't let your drone fly over cities. Not in my country and I suppose not in most places around the world.

My brother however will be happy to use his quadcopter to contribute with some aerial images of fields

Does anyone know a free alternative to Pix4D (the image stitching software)? I'd love to try this myself.

No. An OpenStreetMapper has come up with his own rather less sophisticated pipeline for doing this sort of work:


but it's a way off what this software seems to be able to do.

There's not much magic in what Pix4D does - the algorithms are generally well documented (though tricky to get to work well from what I understand), it's just no-one's done it yet.

This could end up being quite a significant threat to Google Maps.

Unless google maps either hires them or makes their own solution.

I don't mean that their solution will necessarily be better than Google's, but it looks like it could produce top-down imagery that is about as good. Being OpenStreetMap, it will likely be liberally licensed. With the data freely available, the barriers to everyone creating Google Maps competitors will suddenly be minimal. Currently those barriers are very significant (data creation/licensing costs), so them being lowered so dramatically will be unfortunate for Google.

A significant amount of the data is freely available data produced by the U.S. government(s). The issue is processing it into something nice.

The level of coverage that can be done with that data can sort of be inferred from this post:


(they basically processed landsat imagery to get large scale data for the entire globe, and free imagery of the US gets quite a bit finer scale)

n00b question: can anything be done about the mosaicing in the images (look at the corrugated roof, the edge of the concrete etc)?

It's caused by the quick and dirty image scaling algorithm, it doesn't take enough samples per pixel. We used to see this sort of stuff in computer games before mipmaps were invented, which are smaller versions of the image pre-scaled using a fancy algorithm and then used in place of the larger image when it's smaller on the screen.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact