Satellite Image Depth Mapping for Large-Area 3D Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for 3D reconstruction from satellite imagery are inefficient and lack the accuracy needed for large-area coverage due to the lower resolution of satellite images compared to aerial images, and capturing large areas with aerial images is time-consuming or impossible in certain regions.
Innovation Solution
A method utilizing deep learning with a trained Machine Learning Network (MLN) that predicts depth maps from overlapping satellite images, using imaging device parameters and multi-view geocoded ground truth data for training, enabling more efficient and accurate 3D reconstruction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If satellite images are used for 3D reconstruction, then large areas can be covered efficiently, but the resolution is lower compared to aerial images
Solution Approach 1:
A deep learning model serves as an intermediary to enhance satellite image quality. The model takes low-resolution satellite images as input and outputs high-resolution enhanced images, effectively mediating between the limited resolution of satellite imagery and the high resolution requirements for accurate 3D reconstruction
Solution Approach 2:
The patent uses aerial images as training data to teach the deep learning model how to generate high-resolution features from satellite images. The model learns to copy the detailed structures and patterns from high-resolution aerial images and apply them to enhance corresponding satellite image regions
2Measurement precision
If aerial images are used for 3D reconstruction, then higher resolution is achieved, but capturing large areas becomes time-consuming or impossible
Solution Approach 1:
The patent segments the large area into multiple overlapping satellite image tiles. The deep learning model processes each tile independently to generate enhanced high-resolution images, which are then stitched together to form the complete 3D reconstruction of the large area, avoiding the need to capture the entire area with a single aerial survey
3Ease of manufacture
If traditional methods are used for 3D reconstruction from satellite images, then the process is simple, but accuracy is insufficient for large areas
Solution Approach 1:
The patent replaces traditional mechanical photogrammetry processing methods with a deep learning-based image enhancement system. Instead of relying on complex multi-step geometric processing of low-resolution satellite images, the system uses neural networks to automatically enhance image quality, achieving higher accuracy while maintaining relative processing simplicity
Data Source
AI summary
The present disclosure relates to a method for 3D reconstruction from satellite imagery using deep learning, said method comprising providing (101) at least two overlapping 2D satellite images, providing (102) imaging device parameters for the at least two overlapping 2D satellite images, providing (103) 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 (104) 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.