Launch HN: Albedo (YC W21) – Highest resolution satellite imagery
My technical background is primarily in optics/imaging science related to remote sensing. I previously worked for Lockheed Martin, where I met AJ, who is an expert in satellite architecture and systems engineering. We’ve spent most of our career working on classified space systems, and while the missions we were involved with are super cool, that world is slower to adopt the latest new space technologies. We started Albedo in order to create a new type of satellite architecture that captures high resolution imagery at a fraction of the cost historically required. Winston was previously a software engineer at Facebook, where he frequently used satellite imagery and realized the huge potential of higher resolution datasets.
While the use cases for satellite imagery are endless, adoption has been underwhelming - even for obvious and larger applications like agriculture, insurance, energy, and mapping. The main limitations that have prevented widespread use are high cost, inaccessibility, and low resolution.
Today, buying commercial satellite imagery involves a back-and-forth with a salesperson in a sometimes months-long process, with high prices that exclude all but the biggest companies. This process needs to be simplified with transparent, commodity pricing and an automated process, where all you need to buy imagery is a credit card. On the accessibility front, it’s surprising how few providers have nailed down a streamlined, fully cloud-based delivery mechanism. While working at Facebook, Winston sometimes dealt with imagery delivered through FTP servers or physical hard drives. Another thing users are looking for is more transparency when tasking a new satellite image, such as an immediate assessment of when it will be collected. These are all problems we are working on solving at Albedo.
On the space side, we’re able to achieve the substantial cost savings by taking advantage of emerging space technologies, two of which are electric propulsion and on-orbit refueling. Our satellites will fly super close to the earth, essentially in the atmosphere, enabling 10cm resolution without having to build a school bus sized satellite.
Electric propulsion makes the fuel on our satellites way more efficient, at the expense of low thrust. Think about it like your car gasoline going from 30 to 300 mpg, but you could only drive 5 mph. Our propulsion only needs to maintain a steady offset to the atmospheric drag, so low thrust and high efficiency is perfect. By the time our first few satellites run out of fuel, on-orbit refueling will be a reality, and we can just refill our tanks. We’re still in the architecture and design phase, but we expect to have our first few satellites flying in 2024 and the full constellation up in 2027.
The current climate crisis requires a diverse set of sensors in space to support emissions monitoring, ESG initiatives/investments, and infrastructure sustainability. Thermal sensors are a key component for this, and very few are currently in orbit. Since our satellites are larger than normal, they are uniquely suited to capture the long wavelengths of thermal energy at a resolution of 2 meters. We’ll also be taking advantage of advances in microbolometer technology, to eliminate the crazy cooling requirements that have made thermal satellites so expensive in the past. The current state-of-the-art for thermal resolution is 70 meters, which is only marginally useful for most applications.
We’re aiming to adopt the stance of being a pure data provider (i.e. not doing analytics). We...
131 comments
[ 3.1 ms ] story [ 225 ms ] threadOn the timeline scale, if your satellites are only going up in 3 years, is there any reason to believe competitors won’t catch up to your imagery quality?
Btw congrats on trying to democratize access beyond complex sales processes. I hope it works out for you!
Other companies could potentially drive towards higher resolution, but spatial resolution isn't everything. You give up temporal resolution (revisit rates) and overall area coverage, so we expect some companies to continue to cater to the lower resolution type applications.
Some (minor) questions:
> both visible and thermal imagery - at a resolution 9x higher than what is available today
This sentence is infuriatingly ambiguous. Does the 9x refer to only thermal, or also to visible? And what does 9x even mean? Today I can buy worldview3 data which is 30cm/pixel. Are you offering 3.33cm/pixel? I don't think that this is possible without a major, Nobel-prize deserving, breakthrough in optical imaging.
> Today, buying commercial satellite imagery involves a back-and-forth with a salesperson in a sometimes months-long process, with high prices that exclude all but the biggest companies.
This does not match with my personal experience. For many satellites, I can just click a region of interest on a map, and pay the satellite data using a credit card. It is much cheaper to get images from the archive than tasking, but apart from web-UX shenanigans the process is quite streamlined.
> The current climate crisis requires a diverse set of sensors in space to support emissions monitoring
What wavelengths are you aiming for? What spectral resolution? I'd say that most emissions detection happens at the near and short-wave infrared bands, far from the thermal. What climate-relevant emissions are you thinking about, specifically?
I love the contents of this message, but I would like to see it much more specific. This first paragraph it's ok but reads like marketing-speak, not a scientific language.
> While the use cases for satellite imagery are endless, adoption has been underwhelming
This sentence is comically false.
For thermal, we're only doing broadband 7.5um - 13.5um. You're correct in that direct emissions measurements is best made with SWIR. However, there are applications where heat signatures can be used to calculate emissions. For example, if you know the surface material and fuel type of a power plant, you can calculate carbon emission from the heat signature in the thermal image. Climate Trace will use our data for this.
That said, 10cm per pixel is impressive if you manage to pull it through! Cannot wait to buy these images!
They monitor all sorts of things, soil wetness, crop/weed growth, weed identification, cattle location (they used to use helicopters here in Oz for such things)...
Hopefully they have good ML talent on hand that can help sort through the imagery and provide some form of automatic analysis for their customers.
These automatic analyses are much better when they integrate data from several satellites. This seems a job best done by focused third parties that choose the optimal combination of satellites. A single satellite company would be inevitably biased towards using preferrably their sensors, to a detriment of analysis quality.
So no Nobel-prize level achievements needed in optical breakthroughs.
[0] https://en.wikipedia.org/wiki/Intercept_theorem
I've never done it. Can you provide a few links I can try?
A few questions.
How big do you foresee your constellation being?
How fair is the comparison in your Google Photos link? The image on the right seems to have significant artifacting that looks more a side effect of compression.
We used a drone to capture this image at our representative resolution (10cm), but it only captures in JPEG so there could be compression artifacts from that. The image on the right is a 1/3 downsampled version to show 30cm resolution.
Which reminds me, wouldn’t drones always be able to capture better imagery due to lower altitude? It’s my understanding Google maps is all done from aircraft on the higher resolutions.
Use case: following a person in real-time
Real time following (a la Will Smith in Enemy of the State) from satellite imagery remains fictional at present.
The balance of coverage vs resolution is largely dependent on satellite constellation altitude. Higher altitude = more coverage but worse resolution. More satellites also gives better coverage, but adds cost. To get enough coverage for streaming at high enough resolution for tracking an individual person needs a) an infeasibly large satellite constellation b) improvements to image resolution so higher altitudes (and thus better coverage with fewer satellites) c) guaranteed clear skies (for imagery data at least).
When I read statements like this, I tend to read it as "We are not Maxar (Digital Globe)". They are definitely the "legacy" in the commercial satellite imagery market. But what do you see as the discriminator of your company to newer companies providing commercial imagery like Planet and BlackSky? High resolution and thermal imagery are certainly discriminators but other companies are driving that direction too (higher resolution, larger spectral coverage).
I'd like to see your code of ethics.
Like so many technologies today, this could be used by the dark side. You have a responsibility to consider that and put the principles in place to prevent that. Now.
Since much of the tech stack you describe seems to be offered by Planet, (and having experienced the delays and sticker-shock associated with the salesperson-driven enterprise sales process with such companies firsthand), might the industry evolve in such a way that Albedo act as a priceline.com / overstock 'off-brand' reseller of Planet's existing data pipelines? Does developing your own tech-stack basically give you 'table stakes' i.e. signaling a viable alternative to sales channel conversations with existing satellite imagery providers?
I would say that the industry is already trending towards more of a third-party reseller model with companies like Arlula, UP42, Astraea, and numerous others. While we do see the benefits of being able to offer multiple sources of satellite data (virtual constellations combining high-res, low-res, SAR, hyperspectral, etc.) we want to limit our attention to doing one thing extremely well: providing high-res visible imagery.
To your point of developing our own tech-stack for "table stakes" - we certainly hope so. Planet has done a lot of fantastic work here as far as developing robust imagery pipelines but stopped short of removing salespeople from the process. We're hoping to take their approach a step further and remove that manual (human) component from both the sales process and delivery pipelines.
Hope that helps!
I think the 10cm resolution could be as significant in opening up new use cases.
Agriculture is an interesting use case we're also excited about! While it's one of the most obvious applications, adoption has been curbed through historically high minimum order sizes, where cost is an implicit factor. Some providers require up to a 250 sq. km minimum order area, which already prices out a lot of small time farmers. We're hoping to reduce our minimum order area to 1 sq. km. and help enable this semi-neglected portion of the market. One company we're excited about in this space is Enveritas, a company looking using geospatial data to push for sustainability within the global coffee industry.
That's just one I can think of from the top of my head!
[0]: https://monitoro.xyz
What is the best possible way to email you so I can have it and searchable in future to refer back to you (when you are ready for commercial release)?
Here is something I learnt recently. High-Resolution imagery are sleek and nice. This, actually, turns out to be a victim-of-success scenario where potential customers expects such imagery for the sake of it but they won't pay and, honestly, don't need it to be so clear (especially in Agriculture). We realized that the data from a 3-meter resampled is good enough for most needs.
Feel free to fill out the Contact Us form on the website and we'll keep you updated as we get closer towards commercial release!
For a farm sized area, wouldn't an autonomous drone be a lot cheaper&better?
Of course drones are better, as in having an arbitrarily high resolution and being 100% under your control.
For context; we talked to a few tea growers in the eastern side of India. They spend north of $150,000 - $200,000 on drones and UAVs to track/collect data over 20,000 hectares of tea farms. Our sample satellite date can do the same at a fraction of the cost though the precision drops by about 10%. The idea then now is to combine satellite imagery solution + occasional high-precision drone sweeps.
1. https://ag.dji.com/
Anyway, the cost of a 3-meter resolution loaded with multispectral bands and metadata will be just about a dollar per sqkm or even less.
I have lots of small projects I want to work on, but can't bring myself to pay $20-30/km^2 with a high minimum coverage area. Would Albedo help this very small section of the market?
I'd like to continue on these side projects and hopefully use it to build something greater, but the high entry costs make experimentation difficult.
Just curious: have you looked into third-party resellers? They typically have low minimum order sizes and are fairly reasonable, even with a smaller budget.
My suggestion is to have different verticals on the product page and to show specs from their perspective. One for the hedge fund counting cars in parking lots, one for the cops trying to find a missing person, one for the amateur sleuths investigating crop circles.
Quick definition of how revisit rate is used across the industry: it denotes how often a satellite has the opportunity to image an area. So with our full constellation, we'll have a revisit rate of 2-3 times per day for latitudes < 60 deg N/S (info on the website), meaning you could put a tasking order over a certain area for that frequency. Although we expect most of our applications to need something more like a weekly cadence for image collects.
[0] https://www.aaas.org/resources/geospatial-technologies-and-h...
[0] https://www.amnesty.org/download/Documents/12000/asa24010201...
The number of times I want to trace something but can't off Google Maps because of the license terms for using their aerial view, or need a higher resolution version, is becoming far too common.
https://www.space.com/private-satellites-docking-success-nor...
But not so sci-fi.
We’re still in the architecture and design phase, but we expect to have our first few satellites flying in 2024 and the full constellation up in 2027.
AWS has addressed this by making it possible to buy a 3 year compute ‘asset’ which you can record as a capital expense rather than an operating expense.
Pay attention to pricing your product in a way that enables your customers to buy it!
On the AWS topic, we've similarly seen a good amount of reception around a credits-based system.
> Pay attention to pricing your product in a way that enables your customers to buy it!
You're absolutely correct and thank you for the feedback!
> We’ll also be taking advantage of advances in microbolometer technology, to eliminate the crazy cooling requirements that have made thermal satellites so expensive in the past.
You note the parallel with GPS - but this is a 'free spinoff' developed by and for the US govt (and still bankrolled by the US taxpayer) rather than something owned and run by private enterprise. GPS was never developed explicitly with profit in mind. In GPS, the end user also 'owns' their gps traces, in that they can choose to share them with 3rd parties freely. Additionally, there is no value discrimination in gps - all traces are treated the same by the infrastructure owners regardless of how valuable they are (no one is charged more if they use the gps on their phone to enable lots of apps vs someone who only uses it for a single thing). Imagery data on the other hand is much more likely to have different value depending on what is being imaged (there is more value in an image of a busy port than there is of a random patch of empty ocean).
My feeling is that the high res imagery market right now feels more like the market for satellite automatic identification system (AIS) data used in maritime. AIS data is expensive because it is gate-kept by a relatively small number of satellite companies who charge a premium for access and carefully guard how that data is used by onward parties - completely unlike GPS.
As in AIS, because a small number of imagery companies will not only own the infrastructure, but also the images produced by that infrastructure, I'm not sure how it can be commoditised in the same way GPS has been - unless you plan to charge a low flat fee for essentially all images regardless of what they depict and what they are used for?
Tell that to the military. There are people who will pay serious money for a particular patch of ocean if it can be done reliably at an exact time.
Different is more of an over-arching theme, so I'll focus on how we're better and cheaper. We are better in the sense that our native resolution will be 25x finer than theirs - Planet SkySats have 50cm native GSD and our satellites will have 10cm native GSD. This is "better" in the sense that we're more resolute, but we do sacrifice coverage.
As far as cheaper goes, the commercial space industry has made a lot of progress in the past 10 years, and as a result, getting to space has gotten substantially cheaper. This may be slightly oversimplified, but we're hoping cheaper costs for us will transfer over to our users.