Keystone

Inspiration💡

The Keystone project was conceived to address the keystone-related spatial across-track distortion prevalent in hyperspectral satellite imaging. Our team embarked on this endeavor with the objective of minimizing distortion effects, enhancing the accuracy of spectral data, and optimizing spatial coherence in hyperspectral images.

What it does🔎

Keystone operates through a multi-module approach that focuses on rectifying spatial distortions between adjacent spectral bands in hyperspectral satellite imaging. The algorithm comprises three key modules: Initialization, Main Process, and Validation.

How we built it🔨

Utilizing Python and various scientific computing tools, we constructed Keystone, leveraging mathematical models and regression techniques. The algorithm incorporates modules dedicated to different phases of the distortion rectification process, ensuring systematic correction of spatial distortion issues.

Challenges we ran into⚠️

During the development phase, challenges arose in implementing polynomial regression for phase shift calculations, fine-tuning the bisection algorithms, and optimizing computational efficiency in the validation phase.

Accomplishments that we’re proud of🥇

We successfully implemented Keystone, achieving substantial progress in mitigating spatial distortions in hyperspectral satellite imaging. The algorithm’s effectiveness in reducing distortion effects and enhancing spatial coherence stands as a testament to our team’s collaborative efforts.

What we learned🧠

The development of Keystone provided valuable insights into spectral analysis, polynomial regression, and iterative optimization techniques. Our team acquired proficiency in scientific computing methodologies and their application in resolving complex spatial distortion issues.

What’s next for Keystone💭

Moving forward, we aim to enhance Keystone’s precision by refining the bisection algorithms, improving computational efficiency, and exploring methodologies to validate solutions across multiple spectral bands.

Check it out here: GitHub Repo

Detailed Information:

The Keystone approach aims to rectify the keystone-related spatial across-track distortion prevalent in hyperspectral satellite imaging. It operates through three main modules:

Module 1 - Initialisation

This phase establishes an initial estimation by performing the following steps:

• Generating a first guess for subsequent processing using band x+1 that is spectrally adjacent to band x.
• Computing phase shifts between band x+1 and the artificially degraded band x+1.
• Applying polynomial regression to estimate phase shifts between adjacent bands.
• Evaluating local tie points using the SIFT algorithm.

Module 2 - Main Process

This module focuses on estimating the differential keystone by:

• Determining polynomial coefficients close to the first guess via coarse bisection.
• Iteratively computing phase shifts and selecting the best-guess solution.

Module 3 – Validation

This phase involves evaluating obtained solutions for keystone rectification by:

• Applying the inverse solution on band x+1 and computing shifts.
• Performing tie point-related local phase correlation to validate the solution.

These modules collectively aim to rectify spatial distortion, ensuring enhanced coherence and accuracy in hyperspectral imaging.