Planet: Imaging the entire Earth every day | Hacker News Reader

Planet: Imaging the entire Earth every day

nbremer.github.io

464 pointsNaOH133 comments

"Omega tau" made a really nice podcast a while ago about how Planet Labs is approaching Earth observation. That podcast goes into the details how the company and the satellite constellation came to be and what novel approaches they're using. Definitely worth listening: http://omegataupodcast.net/204-earth-observation-at-planet-l...

This is fascinating - especially the talks about the advantages of using COTS electronics.

Once upon a time, the latest and greatest in electronics was produced by military, scientific, and spaceflight applications, with commercial uses being spinoffs that were often up to a decade behind. Now the situation is reversed; consumer electronics are where the innovation is happening, and military/space systems are finding ways to repurpose that technology.

There's actually strong synergy. There's a lot of spin-in and spin-out going on between space and non-space sectors.

COTS and general commodization of "discete units" (parts, subsystems, assemblies, etc.) is one of the biggest drivers for the commercialization of space. It's enabling people to move towards the use of advanced manufacturing techniques, cutting lead time and cost, and enabling batch production.

We recently published an article on how manufacturing lines are changing in the space industry, inspired by best practices in mature commercial sectors like the auto industry: https://blog.satsearch.co/2019-08-08-demystifying-auto-indus...

We see a lot of this happening through our website: https://satsearch.co. We're effectively building a B2B marketplace because of this specific change in the way supply chains are being architected.

Disclosure: I'm one of the co-founders at satsearch.

I think this is only true for small LEO spacecrafts that are more or less 'disposable'. For higher cost GEO and interplanetary spacecraft, consumer electronics are not going to be hardened sufficiently and flight-proven to be used in those applications.

Isn't really about those being higher cost - as the Planet people say, their approach is only possible because they're working inside Earth's magnetosphere.

They also mentioned that they can afford to have satellites fail since they are cheaper and sending up so many. Also won't be possible with one-off GEO birds.

You can access their monthly imagery for free by creating an account at planet.com. I use it to track the progress of infrastructure and other land development projects near me. You can also create time lapse animations of a particular area, which is a super fun way to watch a road or a highway being built.

What is he resolution like? Is it as good as Google's aerial imagery that they show with their maps? (or Bing's)

No way, not even close. At this point Google is doing overhead lidar flights to get incredibly details 3d maps. The Planet imagry is like going back several decades in satellite tech. But on the other hand, it's daily.

Do you have a source for "overhead LIDAR flights"? I was under the impression that data was coming almost exclusively from their StreetView cars (with perhaps a few companies that provide helicopter-based data from time to time). True enough though that Google has a staggering amount of data about the earth.

My apologies. I remembered incorrectly. The aerial flights use computer vision to extract depth maps. While aerial LIDAR is used in GIS, I could not find LIDAR mentioned connected with Google's aerial imagery.

https://blog.google/products/earth/google-earths-incredible-...

Yeah we had a motorhome outside our house and I looked at the 3D map and it showed even though street view car hasn’t visited since 2013. A few months later it updated and showed without the motorhome yet gain without streetview car.

Google imagery is a wide range of 10-100cm imagery. I believe (top of my head) that the planet data is natively 25 centimeters (7-8 band?).

Planet global imagery is 4 band sampled at ~4m (oversampled to 3.0m). Skysats (Planet's high-res satellites) are 0.8m, but image only small areas. (Numbers are based on https://assets.planet.com/docs/combined-imagery-product-spec...)

Nothing is global at high resolution for anyone. Google's imagery is only high res at locations people are likely to look at.

High resolution satellites have a very narrow field of view. You can't image a large area with them. Ditto for aerial. There simply isn't 1m or better imagery globally, let alone 100cm.

I don't think that has anything to do with field of view. It's just that higher res = higher cost so they can't afford to have high res global coverage at LEO.

It really does have a lot do do with field of view. High res satellites fundamentally can't image a very large area (that's how a telescope works). They also usually can't image continuously. (High res means you can't image at the rate you're moving without significant SNR drawbacks - you have to stay pointed the target so that you have a long enough exposure time. You can get around this to some degree, but either way, high-res sats are designed for bursts of targeted imaging.)

You need _way_ more high-res sats than are currently in operation to image the entire land surface of the planet even yearly. So, yes, higher cost, but probably beyond what a company like google could pay. (Think trillions, not billions)

To emphasize on this: WorldView 4, which has ~31cm pan imagery came to a cost of about $850M. And is the size of a school bus. Vs the significantly smaller and less expensive stalites Planet uses. And unfortunately that stallite had a failur and only was usable for a bit over 2 years. High res satellites are really hard still.

Yeah, I got my satellite companies confused. I was thinking of digital globe and their 31 cm offering (got a quote from them fairly recently).

Both Bing and Google use aerial imagery with centimetre-scale resolution (certainly 25cm). PlanetScope is 3m I think, the totally free stuff like Sentinel 2 is 10m (5 day revisit). The best commercial imaging satellites are <0.5m. Below that it's simpler to use aircraft.

3, 5, and 0.72 meter resolution [0]

[0] https://www.planet.com/products/planet-imagery/

I'm pretty sure they've got a 25 cm offering as well.

When I signed up last year I also got a free two week trial period where I could see the daily data. Looks like these days they even let you see what coverage they have of the daily data with a free account, but at very low resolution.

You highlight a region of the planet and they find you all the past imagery of that region. Though most of the daily results are a few strips that pass through the region, with some (or most) of the region missing. They haven't accomplished their ideal of "a picture of the entire earth once a day" but still one of the coolest companies out there.

https://www.planet.com/explorer

I thought they EOL'd the free offering?

They looked me up on LinkedIn using my email address and called by boss to ask me about the service.

I can't believe I've never wondered this before, but I wonder how possible it is to deliberately create a satellite-blocking cloud above one's own property. What height and size it would have to be, etc.

If you could track the satellite could you not shine a laser, or bright light (maybe a mirror reflecting the sun) directly at the camera to mess up the imaging while over your region?

And if an aircraft flies over your property you could get charged with a federal offense (pointing a laser at an aircraft).

Ehhhh, the aircraft flew into the trajectory, are there no other comparable scenarios where there's an intersection that can happen accidentally?

Basically, if you're planning on pointing a laser into the sky, you need at a minimum issue a NOTAM to inform aviators of the hazard. Things like outdoor concerts and some weather observation systems do this.

But... Sending a laser into the sky and intersecting an aircraft that was operating per regulation? That's on the laser-er, and "creating a hazard to aviation" is not a charge that you'll be happy dealing with in many countries.

Doesn't really matter if you're across the globe...

This would be rare and it would only affect the pilot if the laser was tracking the plane as it flew overhead.

This would simply make it very obvious that you are investing substantial resources trying to obscure a location.

Might be fine if all you want is to feel free in your own garden (nude is a common example, but also clothed, if you feel like you might be observed at any time...).

Just make a big warehouse, or put your activities underground. That's what militaries do (also covered drydocks and berthing slots).

Just move to San Francisco. ;)

It's called anthropogenic clouds, and it's possible but requires certain weather conditions to work: you need a source of water vapor that can saturate the air, and a stream of cold air to condensate that vapor. Then it's possible to launch particles into the air that would act as cloud condensation nucleus, and in theory it should start the cloud formation. Anyway, to hide a small area much easier solution would be to have a lot of smoke, or to artificially generate fog which is much easier to do than clouds...

The more common approach is using a large roof, and after that to build where it is usually cloudy.

Yeah, you cant expect privacy sunbathing nude on your own property in 2019.

Kinda scary.

Geodesic dome, cheaper and more durable. As added bonus it protects you from the heat.

One day we will likely dome cities: https://en.wikipedia.org/wiki/Old_Man_River%27s_City_project

One of the uses of a radome (https://en.wikipedia.org/wiki/Radome) is to conceal the appearance of the antenna inside, so that people photographing it can't learn what frequency and direction it targets. That includes by satellite photography.

I don't think it would be very pleasant to have a radome over your house or yard, though, since you usually want sunlight to come in!

For the people commenting on the Firefox+Linux issue: it doesn't work for me, and this is the error I get:

>Refused to create native OpenGL context because of blacklist entry: FEATURE_FAILURE_OLD_NVIDIA

So the real combination is Old NVIDIA drivers + Linux + Firefox. (Maybe not even Firefox, the issue might be inseparable from my old card. I'm not going to try it on Chromium.) The question is, why is my card blacklisted? I think it might be because of a security vulnerability that Chrome hasn't patched.

Agreed, I have an intel embedded thingamabob and it works fine. I wonder if there's a software rendering option in FF that will make it work?

Works for me as well, embedded graphics ubuntu and firefox

> Parts of the landmass can also be missing due to complete cloud cover that day. See the Amazon, Central Africa, or Northern Australia for example.

Aka great places to start a terrorist or drug organization!

The invasion of Iraq got delayed by a few days due to cloud coverage. It would definitely impede surveillance to have clouds often blocking satellites.

This is where SAR imagery comes into play [1]. There are a few SAR constellations online and being worked on that are shifting the discussion about what we can track [2][3].

[1] https://medium.com/from-the-macroscope/its-always-sunny-at-o...

[2] https://www.iceye.com

[3] https://synspective.com

Absolutely great, I am sure we will need these kind of visualizations to survive.

But the planet needs a JavaScript programming competition with the challenge of minimizing CPU load of such fantastic presentations.

It would be so good to have one golden path to browser animations with the lowest energy consumption, one default framework that will be easy to use without deep knowledge of browser quirks and JavaScript optimization tricks.

We are wasting too much time.

Well, I would say, what we need and have is a rich ecosystem of JavaScript applications, which can and does guide both libraries and browser development. I'd say we need, the start-stop and often out-of-step (i.e. careful and experimental) shuffle toward standards. WebGL is just the latest waltz, but that visualisation is built on a bunch of foundational standards: TCP, HTTP, HTTPS, OpenGL, WebGL, etc etc

The computers that got us to the moon where highly efficient integrated circuits. But (I presume) they can't get us to mars.

Personally I'd advocate for messy, but rich platforms, which max out our CPUs in some applications; but give expression to the emergent desires of those who want to tinker.

We don't have time to miss-out on potential wins by working with rigid systems.

Really cool visualization!

Talking about visualizing Earth. Does anyone know of alternative visualizations/projections of Earth (other than globes and maps)?

For example, what would it look like if we could image Earth, entirely, from pole to pole, continuously, like a panoramic picture? (imagine the result as an impossibly wide picture that starts with one pole on the left and ends with the opposite pole on the right)

Thank you.

I don't think that this is exactly what you're asking for, but you can explore most "standard" projections below (definitely try out the first link described below):

- Take a look at [0], click "Earth" in the lower left corner, then try other projections (they appear as the "A", "CE", "E", etc. list). Drag the map around to set the projection's "center". The real time projection calculations are pretty amazing..!

- Go to [1] and toggle all of the categories ("Wall Maps", "Maps of Hemispheres", etc.) by clicking them. The maps are not dynamic / cannot be dragged around (like [0]), but this is a more complete reference list (similar to Wikipedia's list of map projections [2]). The rest of the site is also absolutely incredible (I recommend taking ~1 hour to go through the whole list of projects one by one).

[0] https://earth.nullschool.net/

[1] http://www.radicalcartography.net/index.html?projectionref

[2] https://en.wikipedia.org/wiki/List_of_map_projections

So many questions:

1. How did they get 150 satellites into low orbit?

2. Is it becoming a hazard to have all the satellites and space junk possibly making it difficult to launch anything down the road?

3. Aside from tracking the objects, has anyone made any calculations as to when these things will become a hazard?

They are launched in groups. Lots of them are really small.

https://nbremer.github.io/planet-constellations/

1. They launched by piggybacking on rockets that were launching larger satellites. For instance, they launched 88 on this record-setting PSLV launch: https://www.nasaspaceflight.com/2017/02/indias-pslv-record-1...

2. People generally have a poor understanding of how vast space is. Congestion in specific low-Earth orbits is definitely a concern, but the devil is in the details. For instance, launching to Sun-Synchronous orbit (SSO) [1] above ~600 km is considered to pose significant risk to on-orbit assets, because of how many satellites are in SSO and the kinetic energy their carry, however to really assess the risk, you need to do a lot of detailed population studies (I've done some of this stuff). In a nutshell, launching more satellites can be generally considered to increase risk of on-orbit collision and fragmentation, but quantifying it requires a lot of maths and physics.

3. Yes! There's is extensive and on-going research on this topic. Most people know the Kessler Syndrome [2], but honestly space debris research goes a lot further than that. This review by J.-C. Liou is a great starting point: https://commons.erau.edu/cgi/viewcontent.cgi?article=1233&co...

[1] https://en.wikipedia.org/wiki/Sun-synchronous_orbit

[2] https://en.wikipedia.org/wiki/Kessler_syndrome

1. Many of these satellites are cubesats. They hitch rides as secondary payloads on launches, and in batches. In Planet's case, they've launched many batches, some as large as 20-48 satellites.

2. Not really. Kessler syndrome is the general name for this phenomenon, but it's not a huge concern in LEO. In LEO, there is enough atmospheric drag to naturally deorbit spacecraft. For US based companies, regulations mandate that they have a deorbit plan within 25 years. For most cubesats, their orbits are low enough that they'll naturally deorbit within 5-7 years.

3. Most people generally think that any hazard in LEO will not be caused by launching more objects, but by having objects collide (or having some weaponry intentionally explode objects) causing debris that is spread more widely and too small to track. Currently JSpOC in the US tracks most objects the size of a baseball or larger in LEO and coordinates with various commercial and private entities to put into place plans for collision avoidance where possible.

Thanks for this really concise answer. Here's a paper we did a while back that lists the launch history and some of the different orbit types of the Dove satellites: https://digitalcommons.usu.edu/cgi/viewcontent.cgi?article=3...

I worked for RapidEye during the launch in 2008. Exciting times!

If I remember correctly, their planned lifetime for imaging was 7 years, so it's quite amazing that they're still active 11 years later.

Stupid question: how do they measure speed of the satellite? I mean, on the website they've mentioned the speed of one of their satellite as ~7.6 kmps. Is that the actual speed of the satellite or it's a measure of how much land the the satellite covers per second?

They can measure it by altitude. An orbiting object always has a defined velocity at any point of its orbit. For example, a circular orbit 500km above Earth's surface amounts to velocity of 7.61 km/s everywhere on it.

Makes sense. This is helpful. Thanks, TeMPOral!

The orbital velocity and the ground track velocity only differ by about 6.6% for their 450km orbit, so it's not a big difference. But because they have a relatively circular orbit, you can use the formula sqrt(G*M/R) which gives 7.639km/sec -- so it looks like they are referring to true orbital velocity.

This is helpful, thanks!

Not an expert. But there are a few ways to look at this.

Bodies that are in stable orbit can have only a specific velocity which is a function of its distance from & mass of whatever it is orbiting. So orbiting satellite's speed can be inferred from its orbital height.

There are ground based tracking stations that can technically measure the speed of any orbiting body.

Also read somewhere that LEO satellites could use GPS receivers to get a fix on their position. measuring change in this position also gives velocity. There a bunch of limiting conditions around max-height and velocity for this to work though.

Ah! Got it. Thanks!

Can we receive the signal from these satellites ourselves like we can with the NOAA satellites?

Telemetry, probably; imagery, definitely not. (They sell this imagery, it's not public information like NOAA data.)

Do you mean that imagery can't be received for some technical reason (like being encrypted) or that it would be illegal to save it as it's copyrighted?

The former - encryption is cheap and lost revenue is expensive.

Telemetry and command and control signals are also encrypted.

Here's an article with much more technical talk then I could quickly find on Planet.com's website about one of their first flocks: https://directory.eoportal.org/web/eoportal/satellite-missio... (only tonight have time to read it all)

Anyone familiar the pricing structure of Planet? I couldn't find that on their website.

This paper contains some pricing from mid 2018 (including Planet/Dove):

https://www.researchgate.net/publication/326417596_Benchmark...

this is incredibly cool. Thanks for sharing!

The picture of Monte Fitz Roy is breath-taking. Is there any way to access more of the pictures?

Check out: https://medium.com/planet-stories/earths-wonders-like-you-ve...

Mecca from space: https://www.youtube.com/watch?v=GPgU5OYpwCs

Lot of blind spots (straight stips), are they covered in next satellite revolutions?

I hope they start imaging at least parts of the ocean, like the great pacific garbage patch and other areas of particular interest for meteorologists and environmentalists.

I thought the garbage patch is not visible to the naked eye, mostly made up of microscopic particles, and especially not to a satellite with a 3m resolution?

That's correct, it's something like four particles per cubic meter.

> "fingernail-sized or smaller bits of plastic", often microscopic

https://en.wikipedia.org/wiki/Great_Pacific_garbage_patch

I imagine it should still slightly alter the appearance of the water. Check out "Eulerian Video Magnification" for more.

That work is awesome, but don't they need changes over time in order to achieve the magnification? How would you get changes over time with the plastic pollution in order to view it this way?

Use frames from the same area over a span of time....

What's confusing? There's a picture taken every day.

If you're worried about the time span between the frames, I'm certain that shouldn't matter to the algorithm, it's just more noise and maybe the result takes longer to get..

You can't see the garbage patch from space.

https://www.nationalgeographic.com/news/2018/03/great-pacifi...

I do not see daily updates on planet.com, at least for the free trial account I just created, and for my location (Eastern Europe).

For example, there is a damb nearby which was completed a few months ago, but is shown half-built in their map.

Isn't this like rolling shutter ? But I don't see any stretching of the maps.

Super cool!

I wonder why both clouds and ocean show up as black instead of as actual white and blue imagery respectively.

Are they saving money and bandwidth by not sending back "useless" imagery? Or something else?

Pretty sure its black because they don't image them as its a waste of bandwidth

Most cubesats have small solar panels and can't image all the time (i.e. they need to charge and then image). Mostly likely there no imaging at all over the oceans so the satellites can charge.

(I used to work at planet, this is my recollection of what was the case then)

Usually, for ocean-sized bodies of water, we'd stop imaging entirely to reduce power load and align the solar panels with the sun. Effectively all of the ocean had nothing going on on the surface that we could image anyway.

When Planet gets the imagery, they have a fairly good sense of where it is, but it is still processed to get it to align with ground truth, which is called rectification and relies on matching visible features to known ground features. If the entire image is clouds, this can't work and they don't publish the image because they can't say exactly where it is. Also, there's not really usable information in that imagery I guess.

You could track container ships, in fact they're probably so easy to detect you could run a small computer vision model on the satellite and only transmit those photos.

Obviously the most interesting container ships are the ones that you can see from space but don't show up on marine trackers. Although I don't know how you turn that information into money.

Actually someone must already be doing this, Airbus sponsored a Kaggle competition for detecting ships in satellite imagery https://www.kaggle.com/c/airbus-ship-detection and maybe that's how they caught that Iranian oil tanker last month.

Planet & others are listening to AIS transponder signals from space which I imagine takes out a lot of the need for a bandwidth-heavier computer vision approach. Ships turn off their AIS sometimes when they're doing naughty things, but from what I understand they're usually relatively close to shore when they do that.

You could possibly spot container ships at one instance in a day (the constellation only goes over any point twice a day, ~10.30 AM (local) and 12 hours later (no sunlight)). To what end? You are roughly tripling your power, data storage, and downlink costs, with a loss on your power production (and orbit control, because the Doves use differential drag for phasing) for a poor product that doesn't really have customers.

If you want to track ships that want to be seen, basically everybody has AIS receivers up there (although I understand that's not trivial to sort out from space). If you want to track ships that aren't broadcasting, use radar.

Cool visualization. Would be neat if we could rotate the globe

Really cool tech! I wanted to apply for a job there, but in the end didn't because of their undisclosed military clients.

The visualization also does not show up with the 'Dark Reader' extension on, which is kind of odd.

Welp, the final vestiges of privacy feel like they are being removed.

Can’t see anything on the globe, just a circle.

Portions of the Earth appear as white dots orbit and the Earth rotates underneath. It works in Chrome but not Firefox for me. It's worth spending 10 seconds looking at, but I feel like I wasted some time opening up two browsers to check compatibility.

On mobile it says to use desktop to see all the of the pictures which I can confirm there's way more.

Edit: checking the source code it only shows "20% of the photos" on mobile

    The visual below shows the locations <span class="small-data-note">of 20%</span> of the photos

Can't even see a circle.

"This visualization doesn't work right on Firefox+Linux, sorry! You could try Chrome/Chromium instead."

No Thanks.

Chromium doesn't have all that tracking stuff right. What's wrong with it?

Not true. Chromium talks a lot to Googles servers.

Additionally, "websites" that are made for chrome drive even more people into this monoculture which is bad for the web.

What does it send? Any proof of shady behavior?

Monoculture was bad when it was IE6, it's good when it's google. At least that's the SV Groupthink.

Did not know that.

When Microsoft took the Chromium browser as a base for the new Edge version, they replaced over 50 Google services.

[1] https://www.theverge.com/2019/4/8/18300772/microsoft-google-...

And they replace them with their stuff? You imply foul play but there is nothing in the list.

I think the GP is pointing out that Chromium had G services, period.

So some amorphous and fuzzy evilness by association with zero proof.

It drives the people towards having one browser controlled by a mega corporation. We, at the very least, as developers need to support multiple browsers to prevent the monopolization that could one-day threaten the current style and freedom of the internet that we admire and like (generally).

Seems to work fine here (Firefox 68.0, Radeon R370, Fedora 5.2.7-100.fc29.x86_64, KDE/X-Window). That is, I can see the satellites and the rotating globe with strips of land gradually revealed.

The visualisation works fine in newest Firefox on Linux, though only in Chromium do I get the sound effects (like a high pitched fan noise)

Just so that it's clear to other people: The message is in the text of the page, but it does seem to work on a recent version of Firefox.

I'm on Firefox 68.0.1 on Arch and it's not working.

I'm also on 68.0.1 on Arch and it's working for me, so possibly a plugin is interfering.

Now working for me on Mozilla/5.0 (Windows NT 10.0; rv:68.0) Gecko/20100101 Firefox/68.0. Sats are spinning around a globe, but scanned area does not appear. Works fine on Edge but it's screen reader can't handle the companion article.

This is what it usually means when something is "Chrome-only" - developers like me usually try to stick to standards, but don't have the time or resources to do thorough testing of more than one browser. If it works on Firefox, that's gravy (and my current project does, from a cursory check), but I can't guarantee it in the same way that I can for Chrome.

i.e. this particular dev probably tried it on Firefox/Linux, or got bug reports, and just decided to put up a disclaimer.

Works fine for me in Firefox 69 (beta)

> No Thanks.

Right you are.

Do note that it works fine on my Ubuntu 18.04 machine with Intel graphics on the motherboard, in Firefox 68.

I'm on Ubuntu, Firefox and it's working flawlessly

that's great

Are we moving to the Chrome only world?

Do you mind to share which browser you're using? It's working fine for me on Firefox.

When I load it, the website states that the simulation does not work on Firefox + Linux. Works fine for me when I re-opened it in Falkon (WebKit-based).

Edit: another responder to the parent comment said it works for them on Firefox dev edition on Arch Linux. I'm using Nightly on Arch and it gives the message stated above. With a UA switcher set to Chrome Windows, it loads but renders the image strips on the map as black stripes instead of 'textured' ones. Interesting...

Works for me on firefox-developer-edition. Does not work on vanilla chromium. ArchLinux.

Working fine for me, Firefox 68.0.1 and Ubuntu 19.04

Works for me as well. I love firefox!

Works on Firefox/Android too v68.0

Works fine on UXP (git commit 5a957202b70a0b11078b0cffa8967b63baff5661) / Gentoo Linux.

Using a UA switcher and FF, I can't seem to find anything broken. I think the author of this page is just a bit zealous for chrome.

It also seems to work without UA switcher, but still states it won't work. ¯\_(ツ)_/¯

Perhaps the modern equivalent of “best viewed in Netscape”?

Works in Safari.

Works for me in Firefox 68.0.1 in Win10, kinda slow tho but it might be because of my hardware setup.

Doesn't work for me either. Ubuntu 18.04.3 and Firefox 68.0.1.

The console gives this error: "TypeError: m is null".

Reply on news.ycombinator.com