Thermal Imagery From Space 

This week’s podcast guest is Robin Cole, a senior data scientist at Satellite Vu. With a background in physics, he first got involved in geospatial while working at Surrey Satellites, a company that make Earth observation satellites. Since then, he has delved deeper into the analytics of satellite imagery. He is also active in open-source knowledge communities. Notably, maintains the satellite-image-deep-learning GitHub repository and leverages it to try and educate people on how machine learning can be applied to satellite products to solve a wide range of challenges.

Thermal Imagery Satellites

Satellite Vu is preparing to launch eight thermal imaging satellites over a period of three years. With this constellation, it is expected that the revisit times to certain locations would be as frequent as twice an hour. The sensor will have a resolution of about three and a half meters, backed by high-resolution optical imagery to aid in ground truthing, for instance, to understand whether a linear heat signature is of a road or a pipeline.

Taking thermal measurements of the earth is a useful step in understanding energy use and the built environment.

There are continuously growing concerns that we are using way too much energy, and consequently, a lot of emissions.

Thermal imagery will help in identifying where we are wasting energy, and how changes can be made to effectively reduce wastage.

What Tools Are Required to Work with Thermal Imagery?

Thermal imagery is unique since it is captured from thermal signals. Unlike optical imagery, which can be used with tools that utilize traditional computer vision, special tools are required to work with thermal imagery. At the moment, Satellite Vu has specialized tools developed in-house, which will be used on the thermal imagery they will capture after they launch.

Optical Imagery Vs Thermal Imagery

Optical satellite imagery is captured by very large telescopes installed on a satellite. Similar telescopes are also used on thermal satellites, but the main difference is in the detector and sensor technology used. Thermal sensors measure radiation in the mid-wave infrared band of the electromagnetic spectrum. Cooling the sensor is also required from time to time to minimize spurious thermal signals that may impact an image. These is often from the satellite itself when it is heated up by the sun as it charges batteries with the onboard solar panels.

Thermal satellites can also capture images at night since they detect thermal signals, and do not use the visible light spectrum like our eyes.

They can capture thermal images both day and night, doubling the time they may otherwise be useful. Night time is preferable because the sun is not warming the satellite (and earth) and introducing irrelevant thermal signals into the measurements.

Is Thermal Imagery Affected by Clouds and Shadows?

Clouds can block thermal radiation, but it can penetrate through certain kinds of clouds, particularly the ash clouds associated with fires. Since the ash are very small particles, thermal radiation can get through. In case of fires, thermal (and SAR) satellites can see through the smoke and give real time updates on how the fire line is spreading. It is impossible to see through smoke with optical detectors.

Thermal imagery captured during the day can have some shadow effects as there is some sunlight that is imaged by the detector in the mid-wave infrared range. Even in the case of optical imagery, it can be difficult to figure out where an object starts if it is in a shadow. Since thermal imagery can be captured at night, imaging in the dead of night helps to remove the influence of shadows on the images.

How Is the Spatial Resolution of Thermal Imagery Different?

Spatial resolution is the most talked about metric regarding imagery but it is not a great proxy for data quality! With thermal imagery, it is not always the size of an object that determines its visibility. Because of the nature of a thermal signal, the amount of radiation that is emitted from an object is a function of its temperature. 

An object that is smaller, but significantly hotter than the background would be visible, whereas an object that is larger but has lower temperatures might be difficult to see. 

It is unlike optical imagery where a very small object would be impossible to image as it would be below the possible resolution.

What Are the Applications of Thermal Imagery?

In the built environment, thermal imagery can help identify old properties that are thermally inefficient. 

Usually, the problem for government and property owners is to identify properties that excessively waste energy. Understanding the scale of waste is key to making effective decisions of where to focus limited resources, and make interventions,

like modernizing and upgrading buildings, which is notoriously expensive.

Thermal imagery is also useful in monitoring infrastructure since it can provide information as to what is going on inside a structure. 

Unlike optical imagery which only shows, for instance, the outline of structures in a factory, thermal images can tell whether the factory is operational, and potentially even the kind of processes that are taking place. If it is a chemical manufacturing factory, thermal images can capture temperatures which may be related to specific procedures, and give insight into what is actually being produced.

In terms of tackling urban heat islands, which are highly associated with poor health and living conditions, thermal imagery can help to identify these locations. It provides a way to quantify the magnitude of the effect, advise on where interventions should be taking place, and track whether the interventions made are actually working.

Thermal imagery is also useful in military applications and has been for a long time, primarily because of its ability to capture scenes in the nighttime. This provides a lot of utility for responding to situations that may happen in the dark of night.

Solar farms are also one of the big benefactors of thermal imagery. With strong initiatives to move to renewable energy, more solar farms are being installed all over the world. It is not always obvious where a solar farm should be located or what its impact would be on the environment. 

Thermal imagery can be used to study how a solar farm affects the microenvironment of an area. 

Moreover, since the solar panels themselves age over time and develop issues, thermal imaging can help in making maintenance decisions as it can detect some performance inhibitors, such as cracks and excessive dust.

Is The Time Ripe for Thermal Imagery from Space?

Satellite Vu is going to be the first company to launch high-resolution satellites dedicated to thermal imagery. The big question is why now? Why wasn’t this done earlier? 

The answer to these questions is in legislature changes and economic feasibility. Historically, mid-wave infrared detectors were limited to military and government clients. That has changed in the last couple of years and allowed the free market to enter.

The cost of thermal and satellite technology, particularly the cost of launches, has also gone down. There are several satellite launching companies that have massively reduced the cost of launching satellites into space. 

This has unlocked opportunities for acquiring venture capital to fund activities that would have been unfeasible some time back. Right now, it is cheaper, easier, and faster to launch satellites into space than ever before.

Recommended Podcast Earth Observation Podcasts