Cube satellites equipped with onboard artificial intelligence (AI) can detect fires from space up to 500 times faster than traditional ground-based imagery processing. This technological advancement is enabling Australian scientists to detect bushfires more swiftly than ever before, as reported in a study published in the IEEE Journal of Selected Topics in Applied Earth and Remote Sensing.
Image Credit: Alaskagirl8821/Shutterstock.com
Researchers in remote sensing and computer science have successfully navigated the challenges of processing and compressing substantial volumes of hyperspectral imagery on board smaller, more economical cube satellites, which then transmit the data to the ground for further analysis. This process conserves vital time and energy.
Utilizing artificial intelligence, this advancement enables earlier detection of bushfires from space, potentially before they escalate and generate significant heat. This capability allows ground crews to react more swiftly, helping to avert loss of life and property.
The Kanyini mission, a partnership involving the SA Government, SmartSat CRC, and industry collaborators, aims to deploy a 6U CubeSat into low Earth orbit. This satellite will not only detect bushfires but also monitor the quality of inland and coastal waters.
The satellite is equipped with a hyperspectral imager that captures light reflected from the Earth in various wavelengths, allowing for the creation of detailed surface maps. These maps are essential for applications such as bushfire monitoring, water quality assessment, and land management.
Dr. Stefan Peters, a geospatial scientist at UniSA and the project's lead researcher, explains that traditionally, Earth observation satellites lacked the onboard capabilities to analyze complex real-time images from space. His team, including experts from UniSA, Swinburne University of Technology, and Geoscience Australia, has addressed this by developing a lightweight AI model that fits within the onboard processing, power, and data storage limitations of cube satellites.
This AI model not only reduced the data volume needed to be downlinked to 16 % of its original size but also consumed 69% less energy. Remarkably, it has enhanced the speed of detecting fire smoke by 500 times compared to traditional ground-based processing methods.
Smoke is usually the first thing you can see from space before the fire gets hot and big enough for sensors to identify it, so early detection is crucial.
Dr. Stefan Peters, Senior Lecturer, University of South Australia
To demonstrate the effectiveness of their AI model, the team used simulated satellite imagery of recent Australian bushfires. This imagery served as training data for the machine learning algorithms, which were specifically designed to detect smoke in such images.
Dr. Peters added, “For most sensor systems, only a fraction of the data collected contains critical information related to the purpose of a mission. Because the data can’t be processed on board large satellites, all of it is downlinked to the ground where it is analyzed, taking up a lot of space and energy. We have overcome this by training the model to differentiate smoke from cloud, which makes it much faster and more efficient.”
In a case study involving a previous fire incident in the Coorong, the simulated onboard AI system of the Kanyini satellite detected smoke in less than 14 minutes. Following detection, the system efficiently transmitted the data to the South Pole ground station.
“This research shows there are significant benefits of onboard AI compared to traditional on ground processing. This will not only prove invaluable in the event of bushfires but also serve as an early warning system for other natural disasters,” Dr Peters further added.
The research team aims to test the onboard AI fire detection system in space by 2025, coinciding with the operational phase of the Kanyini mission.
Dr Peters further stated, “Once we have ironed out any issues, we hope to commercialize the technology and employ it on a CubeSat constellation, aiming to contribute to early fire detection within an hour.”
Fire detection from space
The 3D micro-device consists of a modified coverslip and a micro-sphere fabricated by advanced MPL. It can enhance the lateral resolution beyond what is achievable with conventional optics. Video Credit: University of South Australia
Journal Reference:
Lu, S., et. al. (2024) Onboard AI for Fire Smoke Detection Using Hyperspectral Imagery: An Emulation for the Upcoming Kanyini Hyperscout-2 Mission. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. doi:10.1109/JSTARS.2024.3394574