SAR-to-Visible Infrared Image Prediction Under Cloud Cover
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to accurately translate Synthetic Aperture Radar (SAR) images to corresponding visible-infrared images due to differences in electromagnetic radiation interactions with the Earth's surface, making it difficult to derive meaningful information about surface properties, especially in the presence of cloud cover.
Innovation Solution
A method using a conditional Generative Adversarial Network (cGAN) to create a mapping model that translates SAR images to visible-infrared images by training on pairs of matched images, incorporating additional information like surface elevation and sun angle, and employing a loss function that combines GAN loss with Least Absolute Deviations (L1) loss.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Duration of action of stationary object
If SAR imaging is used to observe the Earth's surface through cloud cover, then imaging continuity is improved, but the ability to derive meaningful surface property information deteriorates
Solution Approach 1:
The patent uses visible-infrared images as an intermediary to bridge the gap between SAR images and meaningful surface property information. The neural network learns the mapping relationship between SAR backscatter and visible-infrared spectral responses, enabling translation of SAR data into information that correlates with surface properties without requiring direct visible-infrared observation through clouds.
Solution Approach 2:
The patent replaces direct electromagnetic radiation interaction (mechanical/physical process) with a computational model. Instead of relying on the physical interaction between EM waves and surface properties, the system uses a neural network to compute the relationship between SAR backscatter and surface properties based on learned patterns from training data.
2Measurement precision
If direct visible-infrared imaging is used to obtain surface property information, then measurement accuracy is improved, but imaging reliability under cloud cover deteriorates
Solution Approach 1:
The patent performs preliminary action by training the neural network on pairs of SAR and visible-infrared images before actual use. This pre-training phase allows the system to learn the complex mapping relationships between SAR backscatter and visible-infrared spectral responses, so that when cloud cover occurs, the already-trained model can immediately translate new SAR images into accurate surface property information without requiring real-time visible-infrared data.
3Adaptability or versatility
If neural network translation from SAR to visible-infrared images is implemented, then the ability to predict surface properties under cloud cover is improved, but computational complexity increases
Solution Approach 1:
The computational complexity is shifted to the preliminary training phase rather than real-time operation. During training, the neural network learns the complex mapping relationships between SAR and visible-infrared images. Once trained, the model can perform rapid translations of new SAR images into predicted visible-infrared images and surface property estimates, making the operational phase computationally efficient despite the complexity of the underlying relationships.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to a method and apparatus that can predict the visible- infrared band images of a region of the Earth's surface that would be observed by an Earth Observation (EO) satellite or other high-altitude imaging platform, using data from radar reflectance/backscatter of the same region. The method and apparatus can be used to predict images of the Earth's surface in the visible-infrared bands when the view between an imaging instrument and the ground is obscured by cloud or some other medium that is opaque to electromagnetic (EM) radiation in the visible-infrared spectral range, approximately spanning 400-2300 nanometres (nm), but transparent to EM radiation in the radio-/microwave part of the spectrum. Regular, uninterrupted monitoring of the Earth's surface is important for a wide range of applications, from agriculture to defence.