Help me build a DIY UAV to make small DEMs!
August 29, 2012 9:47 PM Subscribe
I want to put together an unmanned aerial vehicle for capturing aerial photos over a small area (probably <1 km2). Ultimately, I would like to be able to extract a digital elevation model from those photos. How can I get started?
I have read a bunch of stuff over at DIY Drones but I still need some help putting it all together. I know a lot about GIS analysis but not as much as I'd like about remote sensing. I want to build a relatively cheap UAV, but I'm not sure what kind of firepower will be required to get the kind of resolution I need. I need the photos & georeferencing to be good enough to use an image processing service or program (e.g. Drone Mapper) to generate a DEM from overlapping aerial orthophotography that I collect with the UAV.
What is the cheapest way I could do this and still end up with a usable product with high enough resolution to (for example) generate 2 foot contour lines from my DEM?
I have read a bunch of stuff over at DIY Drones but I still need some help putting it all together. I know a lot about GIS analysis but not as much as I'd like about remote sensing. I want to build a relatively cheap UAV, but I'm not sure what kind of firepower will be required to get the kind of resolution I need. I need the photos & georeferencing to be good enough to use an image processing service or program (e.g. Drone Mapper) to generate a DEM from overlapping aerial orthophotography that I collect with the UAV.
What is the cheapest way I could do this and still end up with a usable product with high enough resolution to (for example) generate 2 foot contour lines from my DEM?
Best answer: People are doing this with balloon and kite aerial photography; are you aware of the publiclaboratory people? I don't see it on their website right now but I definitely remember mention of stereophotogrammetry activity in connection with them. Many of their techniques would be applicable to a drone or RC plane/heli as well of course, but kites and balloons are probably cheaper.
posted by hattifattener at 1:08 AM on August 30, 2012 [1 favorite]
posted by hattifattener at 1:08 AM on August 30, 2012 [1 favorite]
(Aha, you probably want their structure-from-motion tag.)
posted by hattifattener at 1:17 AM on August 30, 2012
posted by hattifattener at 1:17 AM on August 30, 2012
i've seen this company's product up close. sorta does what you are looking for, i think.
http://www.uthere.com/
posted by FauxScot at 2:19 AM on August 30, 2012
http://www.uthere.com/
posted by FauxScot at 2:19 AM on August 30, 2012
Best answer: A $50 R/C helicopter will not do this. Not at all. So banish that thought from your mind; you are going to be spending hundreds of dollars on this project as an absolute minimum, and will probably easily break the $1000 barrier. If this is just a one-time need, you might see about finding someone who will let you mount some gear on their airframe and will fly it for you to gather the data.
Most of the drone control systems, e.g. the Ardupilot, will only work with either multirotor helis (tricopters, quads, etc.) or collective pitch CCPM traditional helis. Cheap fixed-pitch helis and coaxials are not supported, at least in any of the software versions that I have seen. And multirotor support is a bit more mature than trad, so I'd probably go that route if you don't have an airframe around that you want to use.
A plane would be cheaper and are generally regarded as easier to fly than helicopters, so that's an option to consider. But you are still talking about hundreds of dollars for an RTF system, and unless you know how to fly already there is a non-zero chance that you'll go through more than one plane as you learn.
I'm not really that familiar with the GIS stuff, or what equipment you'd need in order to gather the data that you want ... although I will warn you that "FPV" systems that allow you to fly the aircraft remotely will add several more hundred dollars to the overall pricetag, and more if you want real-time avionics data overlaid on the video (which you probably do, or else it's pretty easy to get disoriented). Most people use GoPro cameras hooked up to microwave video transmitters, with the GoPro also recording internally for better-quality video after the fact. If that setup will work for you, there's a lot of information around on getting it together. But if you need a different type of sensor, you may have to do more integration on your own.
I'm not trying to discourage you, because it sounds like a neat project, but I think you need to manage your expectations appropriately. This is probably a fairly intensive winter project, and will require some training on your part in order to fly (even with the modern autopilot systems). There may be easier ways to get the data, if that's purely the goal, than building an aircraft yourself.
posted by Kadin2048 at 9:02 AM on August 30, 2012 [1 favorite]
Most of the drone control systems, e.g. the Ardupilot, will only work with either multirotor helis (tricopters, quads, etc.) or collective pitch CCPM traditional helis. Cheap fixed-pitch helis and coaxials are not supported, at least in any of the software versions that I have seen. And multirotor support is a bit more mature than trad, so I'd probably go that route if you don't have an airframe around that you want to use.
A plane would be cheaper and are generally regarded as easier to fly than helicopters, so that's an option to consider. But you are still talking about hundreds of dollars for an RTF system, and unless you know how to fly already there is a non-zero chance that you'll go through more than one plane as you learn.
I'm not really that familiar with the GIS stuff, or what equipment you'd need in order to gather the data that you want ... although I will warn you that "FPV" systems that allow you to fly the aircraft remotely will add several more hundred dollars to the overall pricetag, and more if you want real-time avionics data overlaid on the video (which you probably do, or else it's pretty easy to get disoriented). Most people use GoPro cameras hooked up to microwave video transmitters, with the GoPro also recording internally for better-quality video after the fact. If that setup will work for you, there's a lot of information around on getting it together. But if you need a different type of sensor, you may have to do more integration on your own.
I'm not trying to discourage you, because it sounds like a neat project, but I think you need to manage your expectations appropriately. This is probably a fairly intensive winter project, and will require some training on your part in order to fly (even with the modern autopilot systems). There may be easier ways to get the data, if that's purely the goal, than building an aircraft yourself.
posted by Kadin2048 at 9:02 AM on August 30, 2012 [1 favorite]
I haven't been involved with the RC scene for a little while now, but I'm fairly certain that a $50 RC helicopter will not come anywhere close to meeting your needs here. Firstly, I would write off any helicopter in that price range as a cheap toy (if average prices have plummeted let me know, but I'd be very surprised). Secondly, for the amount of work you're going to be doing, a fixed-wing craft (i.e., plane) is going to be better, from a simplicity standpoint and an energy standpoint. Planes can carry more weight for longer with less energy and fewer moving parts.
On preview, I see that Kadin2048 has already given you a great answer that says a lot of what I wanted to say. I think you need to determine your budget and accurately describe your technical abilities- from there, you can decide how much you want to buy and how long this is going to take.
posted by aaronbeekay at 9:22 AM on August 30, 2012
On preview, I see that Kadin2048 has already given you a great answer that says a lot of what I wanted to say. I think you need to determine your budget and accurately describe your technical abilities- from there, you can decide how much you want to buy and how long this is going to take.
posted by aaronbeekay at 9:22 AM on August 30, 2012
I second taking a look at Public Laboratory and their balloon setup. Great group of folks, and a strong community you can learn from. They sell a balloon kit for $95 but you can also do it yourself with cheaper solutions based on their information.
I'm confident you can do the photo part with a balloon pretty easily. The DEM part may not be so easy, I'm not aware of anyone doing that. Isn't anything aerial photo based going to get the height of the vegetation, not the ground? I did some asking around awhile back about DIY LIDAR. There's decent open source LIDAR software out there but my understanding is the hardware is quite expensive.
posted by Nelson at 10:49 AM on August 30, 2012
I'm confident you can do the photo part with a balloon pretty easily. The DEM part may not be so easy, I'm not aware of anyone doing that. Isn't anything aerial photo based going to get the height of the vegetation, not the ground? I did some asking around awhile back about DIY LIDAR. There's decent open source LIDAR software out there but my understanding is the hardware is quite expensive.
posted by Nelson at 10:49 AM on August 30, 2012
Is the DEM the goal here? For such a small area, you could generate a DEM the old-fashioned way using a surveyor's level and rod, probably for a lot less money (I'm seeing prices on the order of $200-$300 online). It's surprisingly easy to do (though it requires two people -- one to hold the rod, one to take the measurement).
posted by irrelephant at 2:52 PM on August 30, 2012
posted by irrelephant at 2:52 PM on August 30, 2012
Response by poster: I am definitely aware that a $50 RC helicopter is not what I need here! I was referring more to a low-end fixed-wing UAV with as many parts DIY'ed as possible (something like this, but hopefully I could save a few bucks by doing more of the assembly/etc myself. Multicopters are probably out because of the cost, from what I've seen). I should have been more specific about what I meant by "cheap" - I meant <>
I understand that vegetation & houses will show up in any DEM generated from orthophotography, and I've been looking around for a good-enough solution for plots that have little forest cover (for example, an agricultural field). At first glance, it looked like dronemapper offered a service like that, since they will make a DEM for you from aerial photography, but maybe that would still include houses and vegetation &etc. I definitely have more research to do on that process. Forested areas would be the biggest challenge, and I don't expect to be able to survey areas with a large extent of forest cover with this setup. For certain water harvesting applications, though, having a DEM of building surfaces could even be rather helpful, so it wouldn't be all bad if those surfaces couldn't be edited out somehow (and I wouldn't mind doing it by hand if necessary or possible).
Anyway, thanks for all of the great answers! I really love the kite and balloon suggestions, and especially the PLOTS links (don't know how I missed them - now I want a DIY spectrometer too). I am really intrigued by the NDVI imagery they are collecting on their balloons, too - that would be fantastic data to have. Actual surveying is still an option, but I am lazy and it would lack the snazz factor of offering fine-scale UAV surveying services.
If anyone has any further insight on kite vs. balloon vs. fixed-wing aerial photography, I'd love to hear it. A balloon would be much cheaper and easier, but could I still get a DEM from those images? Maybe if there were two cameras on the balloon (stereo imagery), I had enough ground control points, and there wasn't much forest cover?
posted by dialetheia at 12:46 PM on August 31, 2012
I understand that vegetation & houses will show up in any DEM generated from orthophotography, and I've been looking around for a good-enough solution for plots that have little forest cover (for example, an agricultural field). At first glance, it looked like dronemapper offered a service like that, since they will make a DEM for you from aerial photography, but maybe that would still include houses and vegetation &etc. I definitely have more research to do on that process. Forested areas would be the biggest challenge, and I don't expect to be able to survey areas with a large extent of forest cover with this setup. For certain water harvesting applications, though, having a DEM of building surfaces could even be rather helpful, so it wouldn't be all bad if those surfaces couldn't be edited out somehow (and I wouldn't mind doing it by hand if necessary or possible).
Anyway, thanks for all of the great answers! I really love the kite and balloon suggestions, and especially the PLOTS links (don't know how I missed them - now I want a DIY spectrometer too). I am really intrigued by the NDVI imagery they are collecting on their balloons, too - that would be fantastic data to have. Actual surveying is still an option, but I am lazy and it would lack the snazz factor of offering fine-scale UAV surveying services.
If anyone has any further insight on kite vs. balloon vs. fixed-wing aerial photography, I'd love to hear it. A balloon would be much cheaper and easier, but could I still get a DEM from those images? Maybe if there were two cameras on the balloon (stereo imagery), I had enough ground control points, and there wasn't much forest cover?
posted by dialetheia at 12:46 PM on August 31, 2012
Response by poster: Oops, that should say "I meant less than $500, not less than $50". The brackets worked in preview!
posted by dialetheia at 12:48 PM on August 31, 2012
posted by dialetheia at 12:48 PM on August 31, 2012
This thread is closed to new comments.
posted by Ardiril at 10:29 PM on August 29, 2012