Feature Set Dictionary for Satellite Image Translation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Satellite images from different sources often have varying image resolutions, frequency bands, and availability of information, leading to inconsistencies and the illusion of changes in agricultural fields that are not real, making it challenging to generate uniform and high-resolution images for farmers.
Innovation Solution
An agricultural intelligence computer system generates a feature set dictionary from image pairs of different types and resolutions, allowing it to translate images from one type to another, ensuring uniformity and improving resolution by mapping features from one set of images to another, thereby reducing variability and filling gaps in frequency bands.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If satellite images from different sources are used to supplement image frequency, then image availability is improved, but image resolution consistency deteriorates
Solution Approach 1:
The patent creates a reference image from high-resolution satellite imagery and generates synthetic images by applying transformations (geometric, spectral, temporal) to this reference. This copying approach allows frequent image generation without actual satellite passes, maintaining resolution consistency while improving availability.
Solution Approach 2:
The patent introduces a reference image as an intermediary between different satellite sources. By transforming the reference image to match different satellite characteristics (spectral bands, resolution, timing), it mediates between sources with different properties, ensuring consistency while incorporating data from multiple sources.
2Productivity
If low resolution satellite images are interpolated to high resolution grid, then image availability is improved, but detail accuracy deteriorates
Solution Approach 1:
Instead of interpolating low-resolution images, the patent copies and transforms a high-resolution reference image. This preserves the original detail accuracy while enabling frequent image generation by applying temporal, spectral, and geometric transformations to the reference.
Solution Approach 2:
The patent changes parameters of the reference image (spectral bands, temporal timing, geometric position) to generate synthetic images that match different satellite observations. This allows the reference image to adapt to different conditions without losing its high-resolution detail.
3Loss of information
If images with different frequency bands are compared, then spectral information is improved, but image uniformity deteriorates
Solution Approach 1:
The patent uses the reference image as a mediator that contains comprehensive spectral information across multiple bands. By transforming this reference to match the spectral characteristics of different satellites, it maintains image uniformity while preserving and comparing spectral information across different frequency bands.
Solution Approach 2:
The reference image serves multiple functions: it provides the base for generating synthetic images, stores spectral information across different bands, and acts as a consistent reference point for comparison. This multi-functionality allows spectral analysis while maintaining uniformity through the common reference.
4Adaptability or versatility
If satellite images are processed to match different resolutions, then adaptability is improved, but processing complexity deteriorates
Solution Approach 1:
The patent creates a single high-resolution reference image that is then copied and transformed to match different resolution requirements. This approach provides resolution adaptability without the complexity of processing and reconciling multiple source images at different resolutions.
Data Source
Figure 1
Figure 2(a)~2(b)
Figure 3
AI summary
Systems and methods for generation of images of a particular type from images of a different type are disclosed. In an embodiment, an agricultural intelligence computer system receives a first plurality of images of a first type and a second plurality of images of a second type. The first and second types may refer to variances in resolution, frequency ranges of frequency bands, and/or types of frequency bands used to generate the images. Based on the first plurality of images and the second plurality of images, the agricultural intelligence computer system generates a feature set dictionary comprising mappings from features of the first plurality of images to features of the second plurality of images. When the agricultural intelligence computer system receives a particular image of the first type, the agricultural intelligence computer system uses the received image and the feature set dictionary to generate an image of the second type.