The xView2 AI system utilizes similar techniques to object recognition on the ground through each pixel in satellite images, helping rescue workers save time in locating victims in earthquake-devastated areas.
The U.S. Department of Defense is employing a computer vision artificial intelligence (AI) system to support ongoing disaster relief efforts in Turkey and Syria following the earthquake disaster on February 6, which claimed tens of thousands of lives, Interesting Engineering reported on February 21. The AI system, named xView2, is still in the early stages of development but has been deployed to assist in rescue missions in Turkey.
A destroyed vehicle on the streets of Hatay following the earthquake on February 20. (Photo: AFP).
xView2 is an open-source project funded and developed by the Defense Innovation Unit of the Pentagon and the Software Engineering Institute at Carnegie Mellon University. Additionally, researchers are collaborating with several major organizations and companies, including Microsoft. The system employs machine learning algorithms on satellite images to classify damage in disaster areas much faster than current methods.
The xView2 system applies a technique similar to object recognition known as semantic segmentation. This method analyzes each pixel in satellite images and its relationship to surrounding pixels to assess the condition of objects on the ground. The AI then highlights damaged areas in red. Thanks to machine learning support, this method takes only a few hours to implement instead of weeks as before. xView2 is also significantly more effective than traditional methods, which often rely on eyewitness accounts to evaluate damage. Recently, rescue teams have also used drones to survey large areas, but this remains a time-consuming process.
This method is particularly crucial given the number of separate earthquakes and aftershocks that have occurred since the first quake was recorded on February 6. For instance, on April 20, another earthquake with a magnitude of 6.3 struck near the city of Antakya, close to the Syrian border, trapping more people under the rubble.
xView2 has also been used to tackle wildfires in California and assist in flood recovery efforts in Nepal, helping to identify damage from landslides caused by floodwaters. In Turkey, the AI system is being utilized by at least two different search and rescue teams in Adiyaman. Ritwik Gupta, an AI scientist at the Defense Innovation Unit, stated that xView2 can help identify damage in areas that rescue workers may overlook.
However, according to Gupta, there are still some issues to address with the xView2 system. For example, one issue is that the system relies on daytime satellite images, meaning it currently cannot quickly provide data on disasters occurring in the early morning or at night. The service can also be hindered by cloud cover.
