AI-Generated 3D City Maps Using Satellite Radar

In recent days the news is full of information about disasters occurring on our globe. Scientists are wondering how to reduce the impact of such events, coming up with disaster management. However, to create a disaster plan some 3D data is required. Unfortunately, possible solutions for 3D reconstruction have their faults. The lidar method may be too expensive, whereas satellite optical data struggles with cloud cover issues. Satellite radar data instead may be a good alternative for other techniques, especially when it’s combined with AI tools. Thanks to this unique combination outstanding products can be designed, for example 3D City Maps.

From an article in Geo Awesome by Sebastian Walczak.

Presenting SAR2Height

A true revolution was developed by Aerospace engineers at the University of the Bundeswehr Munich, combining radar data and AI. Professor Michael Schmitt and Michael Reclastarted acquired SAR images from German TerraSAR-X and high resolution height maps, based on airborne laser scanning, to carry out a training of deep neural network.

As Schmitt says: “The results were amazing. We trained our model purely on TerraSAR-X imagery, but out of the box it works quite well on imagery from other commercial satellites.” What about the accuracy? This model is able to predict the height component of the building with an accuracy of several meters.

AI solution taming earthquakes

The SAR2Height model is hoped to be a tool for predicting building damages, particularly in case of an earthquake. The vivid example, how useful SAR2Height can be at the present time, is the recent earthquake at an extremely short distance from New York City on 5th of April 2024. It’s a matter of time before we probably see a 3D visualization of the NY’s earthquake aftermath.

For the complete article on 3D city maps CLICK HERE.