Like every organization on earth we use cookies. We use cookies to analyze our product usage. We don't use cookies for commercial purposes. Learn More

How AI and Satellite Imagery are perfect for each other

AI has the power to unlock the potential of satellite imagery. In this article, we provide a practical use-case for AI-powered satellite imagery analysis: Vegetation management for power grids.


The first satellite images of Earth were taken in 1959 by the U.S. Explorer 6. Over the past 60 years, there have been a wide variety of applications for these images, including meteorology, oceanography, fishing, agriculture, biodiversity conservation, forestry, landscape, geology, cartography, regional planning, education, intelligence, and warfare.


Thank you Monsieur Pesquet for the pic

Satellite imagery opens a whole new realm of possibilities when it comes to capturing significant amounts of visual data, and as the sensors improve, the number of potential applications multiplies.

In this article, we’ll showcase the power of satellite imagery by presenting how it can be used to manage vegetation encroachments for power lines, and fight wildfires.

Vegetation encroachment management for power lines: a key application of satellite-AI based applications

Today’s plan to manage vegetation and electrical grid inspections

Utilities and electricity providers must ensure the health and safety of their lines, by regularly maintaining their equipment and ensuring that the environment surrounding the lines is safe. Typically, lines will be inspected and maintained every 5 to 20 years from a mechanical standpoint, depending on the environment and regulations. These inspection and maintenance campaigns are typically carried out by linesmen on foot, or with the use of drone or helicopter imagery.


Utility companies are also required to regularly inspect their network of infrastructures and powerlines to keep them safe from growing vegetation. Not only can falling branches and trees cause power cuts, but they can also ignite - these inspections are not only necessary from an operational point of view, but are also safety-critical.

Critical consequences of wildfires, the case of the Californian electrical utility PG&E

Wildfires have unfortunately become an increasingly regular occurrence around the world. There are a number of possible causes, from stray cigarettes, lightning strikes, and most recently in California, gender reveal parties… and damaged power lines. We won’t focus on human stupidity here. More than 4.5 million homes were identified at high or extreme risk of wildfire in the US alone, with California accounting for more than 2 million of those. In 2017 more than 71,499 fires were triggered in the US. The most destructive wildfires are inevitably due to powerline failures.


The Dixie Fire, started in July 2021 by a tree falling on PG&E power lines, grew to become the single largest fire in California’s history at nearly 1 million acres before being contained by the end of October 2021. DixieFire2.jpg

This, not long after the deadly wildfires of 2018, led PG&E with no choice other than to file for Chapter 11 bankruptcy, to change its board of directors. PG&E has also laid out commitments to a settlement of over $25 billion for all major wildfire victims and groups.


Visual provided by Spacept - Paradise, CA, which was lost to the 2018 wilfires

PG&E’s vegetation management protocol consists of year-round tree clearances using traditional “man on the ground” methods, in correlation with helicopters, drones, and weather stations to assess the risks. PG&E’s Enhanced Vegetation Management program (EVM) promises to reduce the risk of wildfires by inspecting 20% of the highest risk ranked miles and performing approximately 1800 miles of EVM work by the end of 2021.

What if we could prevent the wildfires in California, and everywhere else for that matter?

We recently met up with Spacept, an AI-Automated Satellite Analysis startup based in Sweden as a part of our AI stories series. They researched the Dixie fire and realized that the origin of the Dixie Fire is located in a Circuit Protection Zone (CPZ) with a risk of 11-20%, meaning that the likelihood of PG&E carrying out its EVM program this year in this region is very low. For information, PG&E identified 3,800 miles in this category of risk.

The graph below shows the evolution of the California droughts over the past two decades, from D0 - abnormally dry, to D4 - exceptional drought, with water emergencies and high risks of wildfires.


As the California droughts intensify, as can be seen above, wildfires are more likely than ever. It is crucial that utilities are able to act swiftly to ensure the integrity and safety of their infrastructures.

Spacept, for example, applied their Tree Detector AI model to satellite imagery taken from June 2021 to quickly identify potential hazards and were able to pinpoint vegetation risks.


Origin of the Dixie fire, located by Spacept


The blue path represents cleared sections of the powerline’s path, whereas the orange and red sections represent areas of medium and high vegetation density, respectively.

This is just one of many examples, where satellite imagery analysis can represent a feasible and proactive approach for vegetation management. Of course, it might be used to supplement the efforts of the ground inspection teams and other traditional methods, but it is definitely a promising piece of tech to integrate into the utilities' toolbox to fight against wildfires.

Advantages: Vegetation analysis using satellite imagery

Satellite imagery inherently comes with a number of advantages when compared to traditional inspection methods:

  • Lower carbon emissions (no helicopter and cars used to inspect)
  • Quick lead times compared to traditional methods when capturing large datasets (weeks VS months)
  • Potential for great economies of scale rendering the image capture up to 80% less expensive than traditional methods for large areas.

What sets satellite imagery apart from other capture methods?

Satellite imagery providers like Airbus or Planet can provide images with a resolution of up to 30 to 50cm. Providers typically rely upon a constellation of satellites on pre-planned orbits.


Airbus satellite constellation

Utilities would simply have to specify the location of the infrastructure to be inspected, and the required capture date - The capture date is of course crucial, as the vegetation grows and changes with the seasons.

Providing a time window for the capture date allows the provider to wait until the area to be captured figures on a satellites’ route, however, if the time frame doesn’t allow for it, the satellite imagery providers can modify the orbit of the satellite in order to capture the required area. This option, naturally, incurs higher costs.

In order to build powerful machine learning models to unlock the potential of satellite imagery, the images must first be labeled. This is where LabelFlow can help. LabelFlow is an open image labeling tool that can be used to curate and prepare datasets for machine learning.


There are a number of startups that specialize in AI-Automated Satellite Analysis, developing tools to harness the power of satellite imagery.

Leaders in AI-powered Satellite Imagery analysis


LiveEO offers complete vegetation management solutions using satellite imagery analysis paired with AI capable of detecting high-risk areas. For the first time ever, LiveEO mapped the entire length of the US transmission grid, the largest machine in the world, to determine the length of the grid threatened by vegetation.



Tesselo creates high-resolution vegetation intelligence from low-resolution satellite data, allowing to continuously monitor vegetation with precision and for entire countries.

Their technology is deployed by utilities to plan vegetation management operations for their entire electric grid. Tesselo's forest models are also used by the timber industry for precision forestry, and by public institutions for bio-diverse landscape planning.



Spacept offers a wide variety of Satellite-AI services As mentioned during the example use case, Spacept’s proprietary AI models can locate areas of high vegetation density near power lines, empowering utility companies to improve their vegetation management protocols.


Spacept offer much more than vegetation management - their complete suite of AI-automated tools can detect faulty lights to improve road safety, inspect pipelines using satellite data to detect leaks, prevent losses and protect the ecosystem.


Terrabotics make sense of vast volumes of remote sensing data, from satellites, aerial and ground sensors, at scale. They specialize in all sectors related to natural resources, providing access to superior and more timely insights and analytics – helping to reduce risk, improve decision making, increase transparency and reduce costs.

Terrabotics are also leaders in mine site monitoring solutions, using satellite imagery and AI to deliver key operational, safety, environmental, and social metrics.



Founded in 2016, Preligens focus on providing more informed answers to complex intelligence questions. Their Multi-INT (multi-intelligence) solution is sensor-agnostic, combining satellite and AI to structure data and analyze trends to raise automated alerts, track abnormal activities, and can be used to anticipate the appropriate course of action.
Their solutions are used for:

  • Aerial and maritime surveillance, where their solutions can detect and classify aircrafts and ships, using satellite imagery paired with ADS-B and AIS data respectively to learn and predict trajectories to identify threats on land and sea.


  • PMESII surveillance, or Political, Military, Economic, Social, Information, Infrastructure, which involves a multi-source mapping and monitoring tool for sensitive infrastructures, such as power lines, military bases and ports. Preligens’ AI factory can help users to simulate the impact of abnormal activities and decide on the appropriate course of action.


Featured Articles

View all articles



Get news about our product and releases

© 2022 LabelFlow, All rights reserved.