Hacker News new | past | comments | ask | show | jobs | submit login
Planes Landing in Beijing: Video from Space (mapbox.com)
278 points by bsudekum on March 4, 2014 | hide | past | favorite | 66 comments

They seem to be careful not to show a large object like a plane crossing the "seams" that appear in the video (similar to the seams in satellite imagery on maps where they've patched together data from multiple passes / satellites). When you see the cars on the freeway they seem to disappear at the seams. Does the camera have to be focused on an area the size of the blocks they show? This is still very impressive - I don't mean to downplay the technical achievement, but I'm just trying to get a realistic sense of where the technology is at. I suspect this is video overlaid on static images.

The video field of view is 1.1 by 2 km: http://skybox.com/uploads/10/08/imageryandvideospecsheet.pdf

Purely speculating, not going on anything they’ve said: theoretically when they have more satellites you might be able to patch together a larger area with simultaneous video from several birds. You’d have a slightly ugly angle difference at the seam, of course. There may be technical constraints here that I’m not thinking of, though.

In any case, at 1 m GSD and 30 fps, 1.1 by 2 km is 66 megapixels a second. Even efficiently compressed, that’s a lot of information already.

It looks more like the fov is 0.1 by 2 km (in this particular video).

It seems like the part of the image showing the video is surrounded by pictures (not video). It seems everything is video because he is moving the image around just at the right time. A cool technology enanced with a cool trick.

it's not a trick, it's obviously overlaying video against a google earth background - that's not trickery or cheating, the satellite has a limited FOV and limited bandwidth to tx the data.

"Here you can see Mapbox Streets paired with this video from SkySat-1."

There are some hard, unforgiving physical limits on resolution of photographs from orbit, even with digital correction of the images. For one thing, the camera is necessarily far away, and moving fast at the same time.[1] I was able to follow the specific link just kindly submitted here, which of course shows a rather large object (a commercial airliner). The satellite company's image gallery[2] compares quite favorably in resolution to an earlier company's image gallery we discussed here on HN about three years ago. But it will be quite a while--perhaps never--before some of the statements about satellite imagery that were already being made for propaganda purposes in the 1960s will be anything close to reality, even for classified military intelligence satellites. We won't be reading people's mobile phone screens from orbit ever, probably. Unmanned drone aircraft are at nearer distance and travel at slower speed, and provide a much better platform than earth-orbiting satellites for high-resolution spying.

[1] http://books.google.com/books?id=6DBnS2g-KrQC&pg=PA204&lpg=P...


[2] http://www.firstimagery.skybox.com/

The first link you provided also brings up several interesting implications about continuous surveillance from space. The first is that it is technically possible, but economically difficult. Getting ~5000 high resolution satellites in LEO is not really all that much more difficult than getting 1, it's a matter of scale. But getting funding to put that number up is virtually impossible. Even smallsats and other swarm approaches lack the physical space for the sensors (which is limited by physics, not technology). This is why air breathing drones are becoming so important. Make no mistake, NASA continuous aircraft programs like Helios have ready applications in surveillance. And from those heights, the sensors don't have to be nearly as large as on an LEO bird.

The second is that receiving data from that many birds is problematic. There are only so many receiving stations and only so much EM bandwidth available for communication and recovery of the images. Even switching to drones doesn't solve the problem. Even as of last year bandwidth was a limiting factor for the number of drones up in the air at once. [1] So now we have two problems, getting birds in the air, and we still have the bandwidth problem for downlink.

Finally, analyzing that much imagery is nearly impossible. Automated algorithms help some with certain tasks, but storage, processing, accessing, then analyzing all of that on a continuous basis is pretty much impossible. So even if getting enough sensors deployed is solved, and getting bandwidth is solved, it doesn't mean there's enough people to physically look at and interpret the results anyway. And storing all that for later instantaneous on-demand retrieval is still a problem nobody's really solved.

1 - http://www.defensenews.com/article/20130331/DEFFEAT02/303310...

Bandwidth issues are alleviated by directional antennae, right?

Not necessarily, you still need to collect the transmission. You can still saturate a frequency no matter how focused if enough transmitters are hitting the same receiver.

Exactly. So just deploy enough receivers and focus the transmissions enough so that the frequencies don't get saturated.

Or the receivers track the satellites. This can be done electronically, no need for swivels and servos, just an array.

We've been overcoming hard, unforgiving physical limits for ages.

The obvious solution in this case is to combine images from multiple satellites, ala http://en.wikipedia.org/wiki/Very_Large_Telescope

Yes. It would be unwise to assume that space-based optical interferometry is not currently being done.

The Skybox link has 2 great videos (footage is mostly the same, one is edited with "need more crosshairs"), for those who haven't seen them yet.


youtube/ vimeo direct:

http://www.youtube.com/watch?v=fCrB1t8MncY&hd=1 https://vimeo.com/82644547

http://www.youtube.com/watch?v=BsW6IGc4tt0&hd=1 https://vimeo.com/86196806

I was for some reason completely unaware of this possibility. How much would comissioning such a clip cost, and whats the delay? I looked through the website but couldn't seem to find concrete info. I am surprised that the media isn't using this, like they have taken to using satellite imagery on news shows now. They could show moving imagery of troop movements on the Crimea right now? That seems like it would have a huge impact...

Everyone I know who's worked on microsatellite imaging has talked about the financial district. The thinking goes: watch the Panama or Suez Canal and you're the first one to know when an oil tanker gets delayed, etc..

You can do that right now on a Bloomberg terminal -- definitely don't need satellite imagery. Ships have transponders that tell you where they are. There are definite financial sector use cases, though.

I am more amazed at the lack of smog blocking visibility. Disclaimer: I'm Chinese and I love Beijing as a city, on a good day it's about as clean as LA and on a bad day its...well you can barely see airport terminal when you land.

In LA at least, smog goes to about 5000ft to the inversion layer. Looking vertically through that distance you get much better visibility vs looking horizontally through miles of smog. You can notice the effect looking out a plane window at cruise altitude compared to landing and takeoff.

Exactly. This is why distant city lights often twinkle even more than stars do: there’s more atmosphere between you and them than between you and the stars.

Another effect here is that Skybox’s video uses the pan band, which includes infrared to 900 nm[0], and smog is usually pretty transparent in NIR, depending.

0. http://skybox.com/uploads/10/08/imageryandvideospecsheet.pdf

And a third important factor is that smog is very diffuse (by definition: otherwise you call it a smoke plume), so if you have the bit depth you can just increase contrast until you get a good image.

(I work at Mapbox on satellite imagery, but wasn’t involved with this particular blog post; what I’m saying here is stuff that people in remote sensing Just Know.)

and: "west la" != "central la"

It could be using a wavelengths that go through smog.

That'd be interesting, given this article I saw a while back:


Professor Yang Aiping, an expert in digital imaging with the School of Electronic Information Engineering of Tianjin University and leader of the civilian team, said she was facing tremendous pressure because of the enormous technological challenges.

"Most studies in other countries are to do with fog. In China, most people think that fog and smog can be dealt by the same method. Our preliminary research shows that the smog particles are quite different from the small water droplets of fog in terms of optical properties," she said.

"We need to heavily revise, if not completely rewrite, algorithms in some mathematical models. We also need to do lots of computer simulation and extensive field tests."

Those guys have done a TED talk about designing and launching their lightweight satellites.


Perhaps the most interesting thing about this is the notion that cubesats are not 'toys'. That said, given the low cost of compute one wonders if satellite mapping which is to say generation of structured data on orbit from a high resolution locally obtained imagery is something that might get more work.

Specifically there is a lot of work in the mapping industry for taking a satellite image and turning it into a map. But there isn't nearly as much (if any) work about having a satellite, which can image the same spot repeatedly, providing a downlink of the the map rather than the image. By re-imaging you can just punt on features that are hard to guess and wait for the next pass to see if they will get easier. You might also be able to use multi-angle shadow analysis to pull some 3D feature extraction as well.

This is childs-play compared to what the DoD are up to with WAPS, currently in the format of ARGUS.


Simultaneous video surveillance of 8 x 8km with a single drone, at 15cm resolution.

I'm afraid to say that while this is neat, realtime video maps of the planet will be provided by drone tech, not satellites - for now. When we can stick something in geosync with amazing eyes at a low energy cost (space elevator), this picture may change - but until then... nope.

While impressive, I have a hard time believing this claim:

> As ARGUS floats overhead for months at a time, it dragnet tracks every moving person and vehicle and chronographs their movements, allowing forensic investigators to rewind the footage and watch the activities of anyone they select within the footage.

The amount of footage just one pod collects in a short span of time is monumental. They're storing what amounts to decades of footage for just one day's worth of operation. Maybe they have some criteria that trims unnecessary footage, but I can't imagine that saving all that much space as they don't want to miss anything.

Storage is cheap. If you're storing on HDDs for quick access, the cost is small - if you're storing on archival tape in silos, the cost drops even further.

That is pretty awesome. I would be very interested in limits of modern reconnaissance hardware in orbit, if commercial microsats can produce video with clearly visible cars.

Well, in 1985 US Spy satellites (declassified in 2011) had a resolution of 2-3 feet[1]. "Keyhole-class" (KH) reconnaissance satellites have been orbiting the Earth for more than 30 years and reportedly have a resolution of 5-6 inches[2].



That is precisely why I mentioned modern, in particular would be interesting to know if given a good angle they can resolve phone number being entered on a cell phone with physical keyboard (nokia 3390 for example).

Not sure if troll. The answer is no. 5-6 inch resolution means that one pixel will represent 5-6 inches. You wouldn't even be able to see the phone.

Edit: Sorry, understood your comment now. The real problem isn't resolution, it's "getting a good angle". Unless your target and the satellite's orbit are magically aligned at a precise moment, you're going to have an extremely hard time getting a satellite into position in a reasonable time (the best commercial satellites can do is around 6 hours).

Parent knows this, and was asking what could be done with modern hardware, given that keyhole-class satellites have been up for 30 years.

Optical resolution doesn't follow Moore's law. There will certainly have been improvements, but not like we are accustomed to. Computers can be used to model the atmosphere and attempt to recover resolution, but the returns will diminish quickly.

As long as the laws of physics allow it, its possible.

Not sure why the frantic movement of the map was necessary. Or a looping video, for that matter. Hold still for a while, let us watch stuff moving without moving the map. Reminds me of the faked Roswell "alien autopsy" video that showed a series very quick edits in succession, no way to focus on anything or get a good look at what is going on. Something seems fishy here.

You should be able to see Sean Miller on that northern Africa encampment pretty easily now.

In case anyone is wondering - this appears to be reference to this scene from Patriot Games:


UrtheCast are working on something similar. Their camera recently made it to the ISS and the next step is a spacewalk to install it.


And the third microsat company with operations coming online soon is Planet Labs: http://www.planet-labs.com

Robinson Meyer at “The Atlantic” did a nice overview two months ago: http://www.theatlantic.com/technology/archive/2014/01/silico...

It’s an interesting time in remote sensing.

Ah yes, Uthercast


They are owned by Longford. Longford owned perfect cover for intelligence assets in middle east (oilfields in Iraq), all of a sudden they sold those and took over spy sat startup.

CIA money laundering/cover for intelligence assets type of deal.

Both cameras were installed a couple weeks ago and are functioning: http://blog.urthecast.com/updates/urthecast-cameras-successf...

(Yes, you'd think they'd update their website.)

The near-term future of disruptive spaceflight isn't sending big heavy objects into space---it's sending nano-satellites into orbit around the earth.

Wonder if this can be used over Ukraine/Crimea?

I am pretty certain that's exactly what CIA is doing right now, but using their own spy satellite constellation.

We're still flying U-2s, which, along with certain drones, are far better observation platforms for real-time needs.

It's a mixed bag of emotions, funny, sad, amazed, that the USA is still flying the U2. Which is from the 60's, while the SR-71, the plane meant to replace it has already been decommissioned for over a decade now.

Just FYI, all the U-2s that are flying now were built in the 80s and feature numerous advances over the original generation of U-@s. I agree about the SR-71, I too have a soft spot for that plane. There is rumor of an SR-72 but AFAIK it's still on the drawing board.

Google Maps 2.0

someone is always watching ;)

Paid for with Afghan Heroin.

You wouldn't use drug money for spy satellites, they're happily paid for by Congress. Drug money's for stuff you don't want tied back to the organization - assassinations, corruption, etc.

That's right. see: The Politics of Heroin: CIA Complicity in the Global Drug Trade by Alfred W. McCoy for more details.

This message brought to you by Pat Tillman.

Oh I have a feeling they can do even better than that.

So, I guess the future of Google Maps is where your GPS actually shows your car driving in real time down the highway.

I love living in a video game.

No, it's the future of maps, not Google Maps.

Site is down for me. Any mirrors?

The video from the page is this: https://player.vimeo.com/video/88122560

and says its from Skybox’s SkySat-1.

Ask the NSA how to store data.

would LOVE to have a developer API to skybox. The things one could do!

Urthecast says they'll be offering one, although I don't know what they'll end up allowing: http://www.urthecast.com/developer

What sort of things would you want an API for?

Oh, I was just thinking of how cool it would be to have a live satellite video to create my own mobile apps.

Urthecast seems like it provides images, what would be really awesome is if they offered some short time interval based image refreshing. (acquire new images every few minutes)

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