View source for 20230186561. METHOD FOR 3D RECONSTRUCTION FROM SATELLITE IMAGERY simplified abstract (Maxar International Sweden AB)

=METHOD FOR 3D RECONSTRUCTION FROM SATELLITE IMAGERY=


==Organization Name==

[[:Category:Maxar International Sweden AB|Maxar International Sweden AB]]


[[Category:Maxar International Sweden AB]]

==Inventor(s)==

[[:Category:Tim Yngesj� of Linköping (SE)|Tim Yngesj� of Linköping (SE)]][[Category:Tim Yngesj� of Linköping (SE)]]

[[:Category:Carl Sundelius of Linköping (SE)|Carl Sundelius of Linköping (SE)]][[Category:Carl Sundelius of Linköping (SE)]]

[[:Category:Anton Nordmark of Linköping (SE)|Anton Nordmark of Linköping (SE)]][[Category:Anton Nordmark of Linköping (SE)]]

==METHOD FOR 3D RECONSTRUCTION FROM SATELLITE IMAGERY - A simplified explanation of the abstract==


This abstract first appeared for US patent application 20230186561 titled 'METHOD FOR 3D RECONSTRUCTION FROM SATELLITE IMAGERY

==Simplified Explanation==

The present disclosure describes a method for reconstructing 3D models from satellite images using deep learning. The method involves using overlapping 2D satellite images and their corresponding imaging device parameters. A trained machine learning network (MLN) is used to predict depth maps based on a training set of geocoded 3D ground truth data.

* The method uses at least two overlapping 2D satellite images.
* The imaging device parameters for the satellite images are provided.
* A trained machine learning network (MLN) is used to predict depth maps.
* The MLN is trained on a training set of geocoded 3D ground truth data.
* The depth map of the provided satellite images is predicted using the trained MLN and imaging device parameters.

Potential applications of this technology:

* Urban planning and development: The 3D reconstruction can provide valuable information for urban planners and developers to visualize and analyze the environment.
* Environmental monitoring: The technology can be used to monitor changes in landscapes, vegetation, and natural resources over time.
* Disaster management: The 3D models can assist in assessing the impact of natural disasters and planning relief efforts.
* Navigation and mapping: The reconstructed 3D models can be used for accurate navigation and mapping purposes.

Problems solved by this technology:

* Traditional methods of 3D reconstruction from satellite imagery are often time-consuming and require manual intervention.
* The use of deep learning and machine learning networks allows for automated and efficient reconstruction of 3D models.
* The method can handle overlapping satellite images and accurately predict depth maps.

Benefits of this technology:

* Faster and more efficient 3D reconstruction process compared to traditional methods.
* Accurate and detailed 3D models can be generated using satellite imagery.
* The method can handle large datasets and provide consistent results.
* Automation reduces the need for manual intervention and human error.


==Original Abstract Submitted==

the present disclosure relates to a method for 3d reconstruction from satellite imagery using deep learning, said method comprising providing () at least two overlapping 2d satellite images, providing () imaging device parameters for the at least two overlapping 2d satellite images, providing () at least one trained machine learning network, mln, able to predict depth maps, said trained mln being trained on a training set comprising multi-view geocoded 3d ground truth data and predicting () a depth map of the at provided at least two 2d satellite images using the trained at least one mln and based on the corresponding imaging device parameters.



[[Category:G06T17/05]]
[[Category:G06T7/55]]