You know Albedo best for our 10 cm visible resolution satellite imagery but that’s not all we have in the works.  Taking advantage of the technologies that enable the ultra high resolution visible imagery, we have a special opportunity to also deliver thermal imagery at an unprecedented resolution. Albedo co-collects both 10 cm resolution visible imagery and 2 meter resolution thermal imagery from the same platform, at the same time. The combination of both is something very special.

What is thermal imagery?

Maybe you’ve seen the Predator’s thermal vision from the Predator movies?  Or FLIR (”forward looking infrared”) video from a military aircraft, commercial drone, or a phone attachment? The hotter something is, the brighter it appears in the image. That’s what we’re talking about, from space.

Anything with a temperature emits light (photons), and the frequency (or wavelength) of that light depends upon the temperature of the object. This is governed by Planck’s Law of black-body radiation, which uses Kelvin (K) for temperature units.

Something that is “white hot”, so hot it emits bright light in the visible spectrum? That can be seen in the black-body radiation curve for 5000K (4727°C, 8540°F), peaking at 0.58 micron (580 nm) in the visible spectrum. This is also the same basis for “color temperature” in light bulbs or in white point color calibration, where you might see 2700K, 3300K, 5000K, and 6500K options.

At lower temperatures, in addition to the emitted light being far lower in intensity, the wavelength is also much longer (note the change in the x-axis). For an object at 300K (27°C, 80°F), the peak wavelength of the emitted photons is at 10 microns (10,000 nm). Albedo’s thermal imaging capability is centered about 10 microns in the “long wave infrared” (or “LWIR”) part of the electromagnetic spectrum. Based upon these black-body radiation physics and the everyday temperatures these wavelengths are sensitive to, it is frequently referred to as “thermal infrared”.

Albedo’s thermal imagery lets you see what is relatively warmer or cooler. Look at how much less radiance (light) is emitted at a cooler 280K (7°C, 44°F) compared to 300K (27°C, 80°F) and a hotter 320K (47°C, 116°F). There’s 40% less infrared light coming from the cooler 280K object compared to 300K, and 35% more infrared light coming from the hotter 320K object.

As in visible imagery the difference between a bright white and a dark black is the reflectivity of the surface, in thermal imagery the difference between a bright and dark location in the image is the temperature difference.

Synthetic Albedo Thermal Imagery Background

While we have a drone that we operate with this sort of sensor on it, we know that operating this sensor from space is going to behave differently. Thus we’ve taken a different approach to how we’re showing you what we expect our thermal imagery to be capable of. For our visible imagery examples, we started with aerial imagery. For thermal imagery, we use an entirely physics-simulated process.

We want to be clear about this: this is entirely computer generated imagery, to avoid any confusion about where this scene exists. This is not Hollywood CGI made artistically to wow your eyes, this is physics-accurate rigorously modeled imagery using the validated and widely used DIRSIG toolset developed by the Rochester Institute of Technology.

The scene contains a mix of relevant objects: multiple fighter jets (MiG-25, NATO designation “Foxbat”) in different states of operation, a Boeing passenger plane, a helicopter, an airport surveillance radar tower, a water tower, generators, buildings, and tarmac and grass with different thermal material properties.

The temperature behavior of each object in the scene is modeled, and the emissivity of each material is used to determine the wavelength and intensity of thermal photon emission. These photons are propagated through the atmosphere to the simulated Albedo satellite, where an end-to-end optical and image chain simulation calculates photons at the sensor and through to digital number output, using all of our best estimates for each performance term.

We also use the exact same model and process to generate corresponding visible imagery. We have focused on the radiometric accuracy of this modeling over the photorealism of the resulting visible imagery so that we can ensure we do not misrepresent what Albedo thermal imagery is capable of.

Scene Orientation

First, let’s familiarize ourselves with what this scene looks like in the much more familiar visible spectrum. This imagery has the same image chain simulation and image processing applied as in our simulated visible imagery examples, however instead of starting with extremely high resolution aerial imagery, we start with the DIRSIG computer model and then apply all of those remaining steps to make it as representative as possible.

There are five MiG-25 fighter jets: two parked in the center and right, one pulling out on the left, one emerging from the hangar on the left in the shadows and one taking off (can tell this due to its shadow being further from the plane).

There is a larger Boeing passenger plane in the center. There’s a helicopter on the pad on the top right.

There is a bank of eight generators / fuel tanks in the center. Each is 9 x 3 meters in size, with 2.5 meters of separation between the edges of each horizontally and 1.5 meters of separation vertically.  Center-to-center, they are spaced 5.5 meters horizontally and 10 meters vertically.

There is a water tower on the right side of the eight generators (lighter color), that is bare aluminum. To the left, there is an airport surveillance radar tower, which is mostly thin metal scaffolding / truss structure that is painted black.

The three buildings each have different roof materials and thermal properties.

Most of the scene is grass, with concrete in the middle of the scene (the airport apron or taxiway) and a runway asphalt in the top of the scene which behaves differently thermally than concrete.

Synthetic Albedo Thermal Imagery Example

This is the baseline scene, with all aircraft turned off.

You can see the temperature difference between the grass, apron/taxiway concrete, and runway asphalt (the latter difference being more subtle).

The three MiG-25s on the apron/taxiway read as hotter due to their darker body paint, but their shadows read as cooler.  The shadow of the plane taking off in the top right does not read as cooler because it is moving very quickly and not present long enough to cause the runway asphalt to cool.

The white painted Boeing plane reads as cooler.

The water tower on the right is very hot: think of a bare aluminum playground slide on a hot sunny day. The radar tower on the left is cooler, as it has low thermal mass and conductivity (a truss/scaffolding) and is painted a gray paint that has lower solar absorption.

In the bank of generators, the two on the right are operating, and show as clearly warmer than the non-operating generators to the left.

We have applied a color gradient that helps accentuate temperature differences, where black and dark blues are the coldest temperatures, with purple, magenta, and orange showing medium temperatures, and yellow and white showing the hottest temperatures.

Visible-Sharpened Thermal

The most exciting part of what Albedo is developing is that we are co-collecting this high resolution thermal imagery with 10 cm resolution visible imagery.  This allows us to fuse these products without worrying about complex adjustments for different viewing angles or different times of day between visible and thermal imaging. “What is this hot spot?” is a lot more clear when you have 10 cm visible imagery to lay it over.

Here is an early prototype of combining the two data sources. The detail in the planes, generator bank, and helicopter come across much more clearly in this Visible-Sharpened Thermal product.

Synthetic Albedo Thermal Imagery Further Examples

A powerful aspect of using synthetic imagery is that we control the truth of the scene, and thus can change properties like engines being on, or generators running, and see what that change looks like in the image.

In this image, the engines on two of the MiG-25 fighter jets are now on: the one pulling out on the left, and the one taking off in the top right. There is a clear hot spot present compared to the other jets.

For this next case, the jet engines beneath the Boeing passenger plane are turned on. We notice that our imagery is not quite sensitive enough to notice the change in temperature at the front of the engines, however the very hot exhaust of the jet turbines is heating up the horizontal stabilizers on the tail of the aircraft. This is hard to notice in thermal imagery alone, but when combined with 10 cm resolution visible imagery, the hotter portion of the tail is distinctive.

In another case we ran, we evaluated just the blades of the helicopter being warmer. Further validating that there is a limit to the capabilities of our system, that change was also too small and subtle to be detected in our imagery.

In this animation, we alter the temperature of the top right generator (of the bank of 8) in the simulated scene to see its effect in the synthetic imagery, sweeping it from 320K (47°C, 116°F) to 360K (87°C, 188°F). We have labeled what the truth temperature in the simulation was (this is not intended to represent any temperature measurement accuracy of the system).

A noticeable change is present between each incremental step, illustrating our system’s temperature sensitivity. The top right generator’s heat signature is distinguishable from the bottom right generator, as well as the other six off generators to the left, illustrating our system’s spatial resolution. The change in response across the unchanging areas of the scene is showing the noise of the system.

At high temperatures, we see the automatic gain control of the sensor expanding the dynamic range to handle the hottest temperatures. We are still developing our algorithms to produce output products for your use that maintain dynamic range across such conditions.

Very pretty! But what can it be used for?

Compared to 70-meter resolution thermal available today, without co-collected visible imagery, Albedo is bringing a very novel capability to market with a huge breadth of potential applications.  We have a lot of ideas already for how this might be used, but we’re hoping you can tell us more!

We think there’s tremendous utility for Albedo’s co-collected ultra high-resolution visible imagery and its high-resolution thermal imagery.  Here are a few ideas that have jumped out at us so far:

• Identifying that equipment is hot and running or cold and off
• Locating hot or cold spots on a roof to indicate a heat leak or energy efficiency opportunity
• Indication or measurements of industrial facility energy usage or output, from which carbon emissions, economic activity, and other metrics can be calculated
• Understanding soil conditions (wet soil will heat up differently during the day than dry soil) in order to improve irrigation processes
• Detecting where warmer water is flowing into rivers or bodies of water, indicating water pollution at refineries, power plants, and other industrial facilities
• Monitoring how different parts of a building heat up in the morning for improved characterization of building materials and other property attributes for insurance underwriters

Please reach out and let us know how you might want to use Albedo’s thermal imagery capability!

Click Here to get in touch!

‍