Night and Day: The Power of Thermal Imaging with Satellite Vu

Let's talk thermal imaging! On the #NextSpace Broadcast, we had the pleasure of hosting two guests from SatVu , a company at the vanguard of space technology. Rachana Mamidi spoke with Tobias Reinicke , Co-founder and CTO, and Alex Gow , Sales Director, to learn about Satellite Vu's innovative technology and the evolving downstream ecosystem.

With their recent successful satellite launch, there is no better time to delve into their novel use cases for satellite-based thermal imaging.

This is an abridged transcript of the conversation,?listen to the full conversation on Spotify .

Rachana: Welcome to the podcast!

Let’s start with a little bit of introduction to what kind of thermal imaging satellites, in terms of spectral bands and spatial resolutions, are currently in operation. How much of this imagery is available commercially?

Tobias: Historically, the only commercial system available at the moment that we can get our hands on is the Landsat. And that gives you data of about a hundred-meter pixel resolution, so it's used as a broad area for surveys, etc, but it's quite limited in terms of its revisit.?

There are a number of other commercial systems coming online in the next few years that are really targeting agricultural applications, but these are still quite low resolution. And so if you look at really what's commercially available, it has to be done with either an aerial or drone system at the moment to get down to that building-level thermal measurements.

So there really is no alternative to what we're providing.

One of the biggest challenges as far as I know in building thermal imaging satellites is in minimizing the onboard thermal noise. Are there any other requirements on the satellite bus that are unique to a thermal imaging system?

Tobias: Landsat, like we just talked about, is a satellite that can just hover up imagery as it goes above the Earth. But we have a much smaller, much higher resolution swath.

And because the ground is moving at seven kilometers a second underneath the satellite, we actually need to stop and stare and track the earth as we cover it.

We have to move the satellite itself in space to track the point on the Earth, which is pretty novel. The other one is the thermal detectors need to get corrected or the data needs to be corrected on nonuniformity.

There are going to be weird behaviors in the pixel responses according to what's happening on the ground and the light coming in. That needs to be corrected. Traditionally you can do that by taking imagery of big homogenous areas on the ground such as deserts, ice caps, etc. But we have onboard capability that allows us to capture data which gives us a much more flexible system.

And then finally, we have a satellite that can take imagery day and night.

Obviously, being able to take thermal images at night is going to be really great. If you are doing daytime thermal imaging, it poses its own challenges in that we have a lot of solar reflection issues effectively, so we see a lot of reflection coming back into the spectrum that we're in.

Now that you mentioned onboard capability and a lot of this onboard processing, how does the whole data processing chain, the payload data processing chain work onboard the satellite??

Tobias: So at the moment, regarding the design constraint, we are not in space yet, so the answer to that is it’s not. But when we go to space, the design of the payload processing is that the majority of the data or all the data will be downlinked by your usual satellite ground station mechanisms, and the data processing will be primarily done on the ground, in the cloud.

There is a limited amount of usability of processing on the satellite initially. Typically, the onboard processing is done for minimizing data size, to increase or decrease data latency. We, for better or for worse, aren't a big data company. Our data is going to be fairly limited in size because we are such high resolution and fairly small swath.

So we are perfectly capable of downloading all the data at once. It doesn't mean that we don't want to do more on board.

And whilst compression and pre-processing of the data onboard is going to be something that we will definitely do, the more interesting things are looking at detecting hotspots and even detecting and outlining hotspots.

If you're fighting fires and the satellite can do an automatic fire threshold detection, draw a polygon around it, and send it to the firefighter within minutes, that’s massive.

And that's where we want to get to at some point.

So more requirements for having high computing FPGAs on board then. And speaking of image processing or data processing, Alex, perhaps this is a question for you. How does the whole thermal image processing happen on ground and how do you see this evolve in the future?

Alex: I think the key thing I think we're going through at the moment is deciding what is the data format that our customers most want to use. What the processing level? And with thermal data, does a customer need absolute temperatures? Are they able to use our data in a relative sense for relative temperature? What do they actually need to know about the absolute temperature of an object on the ground??

It's really customer driven in terms of that demand.

Tobias: We need to make sure that the pixels that we capture are on the right place on the Earth so that people can derive insights. The other process that we need to do is calibration on the ground using the data that we get from the space, and then we start processing against a massive database of known materials to create some of these products that we have.

I'm assuming that you are currently focusing on building the thermal payload and not the rest of the satellite. And you are procuring the satellite platform commercially?

Tobias: So we actually have a really good relationship with SSTL, Surrey Satellite Technology Ltd based out of the UK. So the payload on the satellite is their design, procurement, and their build. They have a relationship with a detector manufacturer which they've embedded into the design of their CARBONITE platform. And this is very much the reason why we've been as successful as we have been and have been so attractive to investors is because we're using something that's flown before.

We are keen to see the satellite wrapped up and shipped in May over to Vandenberg to get launched on the 10th of June.

Of course, SSTL has a fantastic track record! Apart from flight heritage, are there any other design considerations when you look for a commercial provider of the satellite bus?

Tobias: We very much go from the customers' and users' perspective, talking about what it is that they want, right? So there's no need to design something for the sake of designing a cool piece of technology. If no one's going to buy that, then you need to get away from it. So we are very much customer-focused. Everyone wants high resolution, everyone wants lower latency. Everyone wants smaller sized.?

We are very much looking at the customers' needs and trying to react to that, saying what is realistic and what is actually going to meet the customer's needs.

Maybe one last question about the commercial satellite platform. In what stages of the satellite development do you wish to participate or do you usually participate?

Tobias: Absolutely everything. For better or for worse, we’re always there.?

At the design review, we put specifications together as to what it is that we require. At the initial critical and preliminary critical design reviews we’re there to make sure everything's going on track. Any adjustments that have been made, we are there to understand the limitations - sometimes the physical, sometimes the timely, and we have to make decisions about what gets done or not.?

So at the end of the day, it's all about tradeoffs and what the customer wants.

Tobias: Oh, absolutely.

We don't need to build the 100% perfect satellite, first off. So let's get the 80% that caters to the requirements and then we have another seven satellites to improve upon. It’s like we’re working in an agile sort of methodology.

Yeah, absolutely. Space is hard and nothing is really perfect in space. So if it works, then that's perfect enough.?

I'm assuming you also fuse optical or near-infrared or SAR or other kinds of imagery with your thermal imagery. And if so, how do you procure these images? And do you see any inadequacies in terms of spectral bands or spatial resolution or frequency of imaging, or geographical coverage in the images of these other bands?

Alex: We are quite lucky in the sense that a thermal image is an optical image. And it's quite simple for someone to look at it and kind of intuitively understand what is hot and what is cold.

But I think the key thing that we have to do is put the data into context.

We've very much taken a first approach, in a commercial sense, to find partners that have other optical payloads where our thermal data could be complementary. There are a number of opportunities whereby we can complement an optical system by filling in that nighttime gap. So you can imagine monitoring something during the day and then filling in the hours of darkness with the thermal sensor.?

So imagine you identify something in your SAR or optical sensors. Perhaps it's a ship that's coming to port. You really want an optical insight. You've got to really get to understand what's going on the outside of structures or physical objects if you want to know, like, is the engine of the ship on or off? That's something that the thermal can help with.

But again, it's all customer-driven, and really it's got to be led and guided by the customers that are going to be using it.

When procuring these different kinds of imagery, do you face any issues or any obstacles or inadequacies?

Alex: I think it’s the usual ones like the cost of very high-resolution data is a very common theme. It's hard to get away from that. I think the proliferation of optical sensors is definitely helping with that. But again, it depends on the application and the use case.

We try to use open source as much as possible. Toby mentioned a few things like improving geo-locational accuracy; we are using quite a bit of Sentinel data at the moment in our production workflow just to improve the geolocation accuracy of our data from a free data source.

And then it is perhaps that some customers might need higher levels of geolocation or attitude and in that case, we might look to procure some more premium data sets to provide that added value.

Tobias: Just to add in, we've seen that if we were to try and consume various radar data sets where some of the providers are still using emails to communicate in terms of tasking and understanding how you get that. So if that's a one-off kind of R&D perspective, then it’s fine. But once you want to put that into a pipeline or into a production pipeline, that's just impossible.

So that is definitely an interesting activity because we're just going through how to expose tasking to our customers via API and we've got to that now. It's good to see what the rest of the industry is doing and to be on some of them are just not ready to be ingested into a production system.

Speaking of high cost and the perceived value of these imagery, have you ever faced any challenges in onboarding, let's say potential customers of thermal data products who have never used satellite-based services before?

Alex: Yeah, a really interesting one. We’re potentially bringing a brand new data source to the market, there’s a lot of customer education that is required because quite often people haven't necessarily thought about how they might use a dataset like this at this resolution, at this kind of frequency, that we are able to provide.

We've had a test platform, we've been flying on an aerial system. Being able to build up this archive of data to provide to potential customers has been invaluable. The second part is that we've had a lot of success with a program that we developed to actually provide that data to our customers.?

Essentially, we can be commercially ready as soon as the satellite launches to start providing data into their workflows and they will know instinctively how is this data going to help solve the problems that they might have.

Can you probably give an estimate, in orders of magnitude, of how much money potential customers of your thermal data products are willing to spend on these kinds of monitoring solutions or insights?

Alex: Yeah, absolutely. That's actually been another great success of our early access program. We have included an agreement for a purchase option, an indication of the size of the contract that they might be willing to pursue. And so far we've got about 57 customers that signed up to our program. That totals about just over a hundred million in purchase options.

The interesting thing with this program is it's not just getting defense customers or just getting commercial customers. It's a really broad range of customers.

Some of the larger interests initially, I think probably on the defense side, but we are definitely seeing a lot of interest in build-to-buy.

Congratulations on your large customer base! That's certainly very impressive. So circling back to the satellite segment, how do you see the typical commercially available satellite platforms fairing in terms of their suitability toward thermal payloads?

Tobias: We've been approached by various payload manufacturers and satellite manufacturers because we have a need and they think that they have a solution. There's always optionality and based on what they're saying, they can do whatever it is that we want them to do. But there's always a question about cost, right?

Our current agreement with SSTL works really well for us. The bus is exactly the right size for us, both in limiting weight for launch costs and agility. So when we look at the sort of bus that we care about, it needs to be agile enough to do the whole movement in space to track the earth correctly and not spend too much time trying to do that, versus not being too heavy, versus having enough redundancy.

Do you see a lot of other thermal payload manufacturers such as companies like SatelliteVu that also might offer thermal imagery analytics? Do you think it's some sort of winner-takes-all scenario, or is there space for multiple players?

Tobias: There's definitely space.

There's quite a big range from where drones and airplanes can work and where Landsat currently is. So anywhere between zero and a hundred meters is fair game.

We are at a certain level, but that by definition precludes us from doing wide-area coverage.?

Whereas some of the other competitors, I’d actually say friends and not really competitors, some of them are 70 meters and are doing really nice wide area agricultural monitoring. Some of them are 40-50 meters doing water, stress, etc. So it's all collaborative and I think it's also going to be complementary.

Alex: You mentioned some of the low-resolution systems being designed for fire detection, but they're not going to be suitable for monitoring the fire once it's been detected and providing real-time information to the firefighters on the ground. And that's where our sensor can come in.

Once you detected the fire, then you have a high-resolution ability to be able to see all the individual hotspots in the fire. One of the things that our sensor has got that’s quite unique is it's actually a video sensor. So it can detect things like speed and motion detection up to minutes worth of video.

That's something that's really piqued the interests of those firefighters - can we estimate the speed and the direction of where this fire is progressing as it grows?

Thank you for a lot of very valuable insights. Tobias & Alex, it's been a really nice session with you guys and hope at some point in future we could also be able to collaborate and do something fun in the thermal imaging space.

Tobias: Thank you for your time!