Epipolar Image Alignment for Satellite Depth Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems struggle to accurately align and superimpose satellite images captured from sensors with different line-of-sight directions, leading to positional shifts and challenges in depth detection, particularly in distinguishing cloud formations.
Innovation Solution
An image processing device and method that aligns image data from sensors with different line-of-sight directions using a geodetic system to eliminate positional shifts, generates epipolar plane images, and detects depth from streak patterns in these images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image data from sensors with different line-of-sight directions is superimposed without alignment correction, then the processing is simpler, but positional shifts occur and depth detection becomes inaccurate
Solution Approach 1:
The patent applies preliminary action by performing alignment processing before superimposition to eliminate positional shifts. The system pre-calculates and applies transformation matrices to correct the positional deviations caused by different line-of-sight directions, ensuring accurate overlap of image features before depth detection is performed.
Solution Approach 2:
The patent introduces an intermediary approach by using a virtual reference frame and transformation matrices as mediators between the different sensor coordinate systems. These intermediaries facilitate the alignment of image data from multiple sensors without requiring direct complex coordinate transformations between each sensor pair.
2Measurement precision
If cloud distinction is performed using only reflection intensity from visible images, then the method is simpler, but cloud and ground features cannot be reliably distinguished
Solution Approach 1:
The patent merges multiple data sources including visible image reflection intensity, infrared image brightness temperature, and depth information from epipolar plane images. By combining these complementary measurements, the system achieves more reliable cloud distinction, using infrared data to identify thermal characteristics of clouds and depth data to determine vertical positioning.
Solution Approach 2:
The patent creates a composite analysis approach by integrating multiple types of image data and processing techniques into a unified cloud detection method. The system combines spectral information from different wavelengths with geometric depth information to form a comprehensive characterization of cloud features that cannot be obtained from any single data source alone.
Data Source
AI summary
Provided is a method of acquiring image data generated by imaging by the imaging means arranged to have different line-of-sight directions, superimposing the image data while performing alignment in such a way as to eliminate absolute positional shift caused by the different line-of-sight directions in each piece of image data, selecting a transverse line that transverses an overlapping portion of the superimposed image data according to a direction in which ranges of each piece of image data in a superimposed state are shifted, generating epipolar plane image data by arranging each piece of image data of a portion where the selected transverse line and the overlapping portion overlap in an order determined by magnitudes of inclinations of the line-of-sight directions corresponding to each, and detecting a depth of a subject appearing in the image data from a streak pattern appearing in the generated epipolar plane image data.