Solo for Egyptology

[Kevin Cain]

It's really useful to photograph walls and columns at Egyptian temples in a fairly dense grid; we've built a custom 12m pole that allows us to do that, as seen in this picture:
http://www.insightdigital.org/entry/images/ramesseum/ram6.jpg

My question is whether a 3DR drone such as the Solo can provide the kind of reasonably regular photo spacing in this image:
http://www.insightdigital.org/team/images/9/91/Litanies_North.jpg
(Note: I'm talking about the grid of camera positions close to the wall, not the other views farther back.)

As many may have guessed from the images, we require this tight grid and overlap to ensure that our image-based 3D reconstruction code has the input it needs. If the flight path wavers a bit, we'll lose the coverage we require and the reconstructed model will end up with a void in that spot. The flight paths don't have to be absolutely perfect (and what I show above illustrates that), but they have to be pretty good.

My feeling is that this sort of linear grid shooting style calls for autonomous, programmatic flight paths -- not a pilot flying via the Tower app, et cetera. I imagine that even more true if we're shooting a cylindrical grid around a column, as with the camera positions shown here:

Any opinions?

A few notes/questions follow.

0. We can't rely on GPS in Egypt. GPS for civilians was banned in 2003, restored in 2008 -- and I can't find any mention of its availability right now. Beyond this, we can't count on GPS at the temple we work at, because the stone temple itself blocks the signal to the satellites. Solo wants to use GPS, but there seem to be manual modes: Advanced Flight Modes | 3DR | Drone & UAV Technology. Could we get the controlled photo grid I showed above out of a Solo in a manual mode?

1. The kind of grid shooting I noted above requires good station-keeping, obviously -- compensation for slight wind, et cetera. Has anyone done any work on station-keeping for the Solo? If we were using a Phantom 3, we might be able to use their camera-and-sonar tracking to keep oriented without GPS. On the Parrot AR-Drone line there have several open source station-keeping projects (e.g.: tum_ardrone - ROS Wiki). Any equivalents for the Solo?

2. The PCL/ROS community has spent several years working with the AR platform and has some nice autonomous flight results. (Like the Raspberry Pi, good software can sometimes trump limited handware.) There are many examples of universities that have written interesting papers around this architecture -- see these examples (and ROS distro) above (tum_ardrone - ROS Wiki), or this researcher: Computer Vision Group - Jakob Engel.

Academic AR work has shown good progress in autonomous flight, which as you'll gather, I consider much safer and more effective for image-based modeling than manually flying a drone and trying to manage to get complete coverage for a complex set of surfaces. A few examples:
Autonomous Parrot API: AutonomyLab/ardrone_autonomy · GitHub
Navio drone autopilot: NAVIO+

I realize that the Tower project is a similar open initiative with lots of promise; but I don't see anything applicable to our needs in that project yet. Sure I be looking somewhere else, or am I reading this wrong?

Thanks for any contributions!

 

[Kevin Cain]

Would a GPS repeater be useful to redirect signal into the covered areas of the temple I mentioned above? Here's a photo to give a sense of the interior/exterior 'perforated' spaces we'd like to fly:

Does anyone have any experience or advice with this equipment?

 

[NigelTM]

Solo relies heavily on GPS for station keeping. Even then it is only so accurate and can misreport it's position even if holding a stable position. If GPS is not available/capable then you will have issues station keeping and tagging the images with a 3d coordinate position or bearing for the image processing.

The ARs and similar drones use a down facing camera and image processing to hold a stable position. However,this does not give/use a coordinate system/reference, which you would need to stitch together the images for the 3d model.

Simply put, if you need reasonable high accuracy coordinates/location to place the images you will need either a high accuracy GPS or some form of locally generated coordinate system. Station keeping can be managed using a down camera type system (which I believe is in the Made For Solo production queue) .

While I think Solo is very capable and fun to fly as an image/video platform I don't think it is the platform for this type of high accuracy/fine detail image capture.

 

[Kevin Cain]

Thanks Nigel, I agree that GPS will naturally drift quite a bit. Thankfully, for me the question of reported GPS accuracy is not an issue. My 3D reconstruction code (like most) doesn't require geolocation input. Instead, again like most, I compute the camera poses with multi-view stereo methods; in this way, the model of the single column I included above did not require any input beyond the images (with EXIF camera extrinsics) themselves; the positions are recovered with a good deal of accuracy.

What I'm wondering about is the ability to fly the Solo on paths that approximate the photo 'grids' I've obtained with our pole, above. I don't have a sense for how much drift we'll see moving up and down fairly close to the large walls and columns I showed in the temple image above.

We can fly arbitrarily slowly, use GPS repeaters to help stability, any other tricks people might suggest, et cetera. But perhaps the reflected wind from the Solo itself will make it impossible to fly in straight lines (+/- 6") if we're ~1m from the walls or roof slabs.

I'm hoping that some of the experienced folk in this forum will be able to chime in with 'sounds reasonable' or 'doubtful', as I don't have any experience flying Solo.

 

[John Oster]

I would say doubtful but that is based not so much on prop backwash but wind factors in the vicinity. I've been to Cairo (see avatar image of me) and know that it can be very windy at times. That would wreak havoc. The Solo gives a stable video performance, but that is due to cooperation with the gimbal. I've thought about building a device that uses the Solo gimbal by itself. (Yuneec makes a handheld that also attach's to their drone.) Traveling along a horizontal wire you could get fairly accurate imaging. You could robotically lower the gimbal unit and continue making passes back and forth as you descend on the wall. But alas, a crew would need to re-stage the wire every time you have completed a wall.

 

[Kevin Cain]

Thanks, John, good points all.

I've worked at various sites in Egypt and you're right -- the wind can be a mighty force. At quarry sites where we've done kite photography, that's actually a plus. For a quadcopter, not so much. In Luxor we have pretty reasonable conditions; over the years we have had wind only very occasionally.

My fear is not so much the wind but the intrinsic drift in single-axis flying.

The wire system you describe is one we've considered ourselves -- a much lighter version of the aluminum pole we already use. The trouble is that we can't touch the walls of the temple, so the wire frame would need to be freestanding, like a loom or 3d printer for houses. The walls are not that large, but at ~5m x 10m each, the tension for the wire would have to be awfully high to prevent sagging as the 'shuttle' moves towards the string center.

I see you're in San Francisco so we should meet at some point!

 

[John Oster]

Speaking of S.F., I am a retired stagehand and I have primarily built film scenery for the last 39 years since I started in 1976. We occasionally do films on Alcatraz. We are not allowed to touch the walls inside the prison either. We do a lot of wedging. That's where we developed the phrase "Let me tell you about our two friends, gravity and compression!" I've worked with special effects quite a bit and have seen some innovative solutions to unique problems. If you have two stout poles, you only use them to hold the cable in the air. You run the cable past them on both sides if necessary and tie them to a water truck if you have one, or any old military tank you have lying around! Presuming you have ratchet straps, you ratchet that cable as tight as you can get it. But if you don't have vehicles, you do have a lot of sand don't you? That works great too! Good luck with whatever you come up with. Sounds like you know what you are doing. I was there in 2005 working for Intel Corp.. (Head Carpenter) When the uprising in 2009 happened I wondered what happened to Dr. Hawass. Are you familiar with him? I think his title was Head of Antiquities.

 

[Kevin Cain]

John, everything you say would work well for us and reminds me of a lot of the 'practical' options I've been thinking of in the past months. Like you I have a background in theater and film and know how much you can do with a small team who knows what they're doing. I have no doubt that if you were coming along, it would be easy enough! I invited a very good IATSE stagehand to come to one of our projects in the Yucatan (see our site if you like: www.mayaskies.net) and he could move mountains also single-handed.

The trick here is that we have a large number of moves for any rig we might use, and we need to fabricate and fly in everything (including tools) with us, as it's difficult or impossible to do so on site in southern Egypt. Add to that our almost non-existent budgets (just buying a Solo is stretch for the team) and a team that isn't used to this, it's hard for us to rely on the same sorts of techniques that we could readily use on set.

Over time I think it's clear that drones or one kind or another will be robust enough to handle tricky photo assignments like this. It may well be that they are just 'not ready for prime time' for these specific uses, as people have been suggesting in this thread. If that's the case, then I'd love to return to you as we think out the options for a wire rig, or similar.

But for now I do think that the Solo (with Tower) has a shot at giving us what we want. So far I haven't seen anyone present a compelling reason not to try.

And yes! I've met Zahi many times (most recently, last month) and presented for and to him in his massive Cairo office on one memorable occasion.