Geospatial Polynomial Transformation Toolbox

Overview

The Geospatial Polynomial Transformation Toolbox provides a robust framework for geospatial image transformation, resampling, and interpolation using polynomial regression. This toolbox enables users to perform coordinate transformations, pointwise corrections, resampling, and piecewise regression to improve georeferencing accuracy.

---

Dependencies & Installation

1. Clone the Repository

To ensure all required submodules (including the Genetic Algorithm (GA) regression module) are included, clone the repository recursively:

git clone --recursive https://github.com/HesamTaherzadeh/GeometricRegressor.git
cd GeometricRegressor

2. Install Python Dependencies

Ensure that the required Python packages are installed:

pip install numpy matplotlib PySide6

---

3. Install and Build the Genetic Algorithm (GA) Submodule

The Genetic Algorithm (GA) module is located in thirdparty/GA. Navigate to the submodule and follow the build instructions:

cd thirdparty/GA
mkdir build && cd build
cmake ..
make

After building, you can run the GA binary as follows:

./genetic_algorithm ../config/cfg.yaml

To run tests for GA:

./runTests

Python Setup for GA

To use the Python bindings for the Genetic Algorithm, follow these steps:

cd ..
python3 -m pip install pybind11 numpy
python3 setup.py build
python3 setup.py install --user

To verify the installation:

import genetic_algorithm as ga
print("Python bindings successfully imported!")

---

Key Features

• Polynomial Regression for forward and backward coordinate transformation.
• Pointwise Interpolation using Multiquadratic (MQ) and Local Distance Weighted (LDW) methods.
• Resampling with bilinear interpolation and multithreading for efficient processing.
• Piecewise Regression, enabling adaptive transformations for different spatial regions.
• Graphical User Interface (GUI) for interactive visualization of transformation results.

---

Modules & Documentation

Each algorithm is documented in its respective file:

1. Polynomial Regression

   • Implements forward and backward transformations using polynomial equations.
   • Forms the basis for all other modules.

2. Pointwise Interpolation

   • Computes localized displacement corrections using MQ and LDW interpolation.
   • Enhances ICP coordinate accuracy.

3. Resampling

   • Performs image resampling with polynomial transformations.
   • Uses multithreading for optimal performance.

4. Piecewise Regression

   • Allows dataset splitting into multiple regions based on user-defined lines.
   • Conducts localized polynomial regression for improved accuracy.

5. GA

   • Chooses the Best terms to avoid overparameterization error
   • Performs two meta-heuristical search for terms for both backward and foreward

6. [Project management]

   • Defined a project class that saves all of the neccessary information processed into a .kntu file, named after KNTU university
   • This utility allows users to save the project and their terms.

7. UI Tutorial

   • A set of GIFs to showcase the power of the code

---

Usage

1. Load an Image and Ground Control Points (GCPs).
2. Perform Polynomial Regression to compute transformations.
3. Apply Pointwise Correction to refine the transformation accuracy.
4. Resample the Image to generate a transformed grid.
5. Use Piecewise Regression to analyze spatial variations.
6. Visualize results interactively with quiver plots.