Hacker News Reader

Comment by csteubs | Hacker News Reader

I'd argue another point; satellites are a major bottleneck to GIS as we're familiar with. With only ~300 or so commercial imaging satellites in orbit, current revisit rates are sufficient for most use cases. Maybe one or two images/day for X point depending on weather. Low earth orbit is getting crowded, fuel to raise and lower orbit is limited onboard, expensive maintenance cycles, etc.

I'm working on a startup trying to solve this problem now. The idea is to "crowdsource" aerial imagery using a network of camera arrays on 80,000 commercial flights/day. We stream the video feed back to earth using in-flight wifi and overlay the visual diff to create a map that shows minute-level changes. There are huge built-in cost savings that come with not having to develop, launch and maintain a satellite constellation, and with a vastly higher revisit rate (as high as 4,532/day in some areas). We're at notasatellite.com

View on HN

Streaming the video feed using the onbord wifi? I think you are overestimating the quality of the Internet connections available on flights.

Having worked on inflight connectivity apps and Earth Observation respectively, for a couple of startups I'm inclined to agree overall. Workarounds would be:

[i] ground sync, which would be fine for a lot of use cases where higher revisit rate and sometimes less cloud cover is sufficient to make the imagery valuable, and open up non-wifi flights. Ground sync can be over regular 4G or even swapping flash drives [some W-IFE boxes update this way, though infrequently].

[ii] Getting dedicated bandwidth from the satellite/a2g internet provider, who can potentially allow you a lot more than is made available to an individual consumer, but obviously at a hefty charge.

[iii] using streaming codecs designed for low kbps rates - though this is obviously lossy, and pixel perfection matters for a lot of EO use cases.

Great points! We're incorporating aspects of all three workarounds. Ground sync is already possible (wifi isn't available on many regional short-hop flights so we eat the latency), and we're discussing dedicated bandwidth with in-flight wifi providers. We've been paying the $14.99 market rate for customers testing the MVP who receive free in-flight wifi for their efforts, so the resulting cost would be much cheaper. To your last point, we load the last-known tiles expected to be captured on the flight in order to take a visual diff in the air, and then send that diff back down to earth instead of the entire image. This results in a roughly 80% savings in data but requires more hardware.

Yes and no. Connectivity tests have been an important part of the early work so far but depending on the airline and route, we've consistently averaged 6Mbps upload pre-Covid which has been sufficient. Passenger volume is significantly down post-Covid and that upload number has been about 3x lately. Wifi coverage is ~100% over land and the North Atlantic, so reliability of service is a negligible issue. American Airlines has recently upgraded their in-flight wifi to Gogo's 2Ku systems and we've hit 15Mbps up on those flights, but as another commenter mentioned, the plan is to carve out dedicated bandwidth onboard the flight for better connectivity. A connection is only optional, as we save the recorded data and can upload once the flight lands using wifi on the ground (much cheaper, but with a few hours of latency between recording and becoming a map tile).

I'm clueless in this space but I'd imagine the cost of adding these sensors(installation, ongoing maintenance, certification etc) to airplanes is non-negligible? I assume you're targeting small personal aircraft for the sensor array and not some airline with a fleet of 737s?

The sensors are mounted internally to the window in order to avoid any sort of external retrofit work, tunnel testing, FAA approval, etc. The MVP is actually just an app that uses the phone's camera to the same effect. We're talking to a number of carriers and exploring all the other deployment options as well (including a frequent flyer revenue share program).

What's the value of having the image be sent and seen instantly versus in a couple of hours later when the plane lands?

The value really depends on the use case, but having the image sent instantly enables finer temporal resolution. Not all flights will have working wifi so all the devices (including the app) can save the data and upload once the flight has landed.

The flights per day/daily revisit rates are pre-Corona, yes?

They are, but that number represents U.S. flights only and is pretty grossly underestimated. Flight volume hasn't changed significantly, but the passenger volume is way down which is a good thing for our purposes. There were 64,523 flights on March 29th (around the peak of the initial wave of lockdowns), compared to 154,000 back in January of this year. I'm using a dumpster fire of a source (Forbes) for those numbers but they can be found here: https://www.forbes.com/sites/gabrielleigh/2020/03/30/the-num...

Even if global flights slowed to post-2001 levels, we'd still be capturing vastly more data than the ~300 or so commercial imaging satellites. Our margins are fairly wide in that sense.

Reply on news.ycombinator.com