geocoreg

A library to simplify the process of co-registering geospatial data. Althought some great libraries provide co-registration functionalities, such as scikit-image and kornia, they don't directly support geospatial data. This library aims to provide a simple and easy-to-use interface to co-register geospatial data in a pythonic way. The library design was created to allow the user to select different co-registration methods from different back-end libraries. For an easier utilization, Xarray DataArrays are supported (dask is not supported yet).

Warning: This library is still in early development and the API may change.

Installation

To install the package, run the command below

pip install git+https://github.com/lbferreira/geocoreg

Usage

To coregister a single image or multiple images (e.g., a time series) in relation to a reference image, you can use the coregister function, as shown below.

import xarray
import rioxarray
from geocoreg import visualization as vz, xr_registration

# Load sample data
da = rioxarray.open_rasterio("naip_example.tif")
da = da.assign_coords(band=['red', 'green', 'blue'])
# Simulate a time series with some shifts
shifted_series = xarray.concat(
    [
        da.assign_coords(time=0),
        da.shift(x=10, y=10).assign_coords(time=1),
        da.shift(x=-5, y=8).assign_coords(time=2),
    ],
    dim="time",
)
# Export a gif with the original series
vz.export_animation(shifted_series, 'time', file_name='original_series.gif')

# Define the reference image as the first image of the time series
dst_img = shifted_series.isel(time=0)
# Coregistrate the time series with the reference image
ts_registrated = xr_registration.coregistrate(shifted_series, dst_img, registrator='pcc', registration_band='red')
# Export a gif with the registrated series
vz.export_animation(ts_registrated, 'time', file_name='registrated_series.gif')

Acknowledgements

This library was developed as part of my research work in the GCER lab, under supervision of Vitor Martins, at the Mississippi State University (MSU). This research is funded by USDA NIFA (award #2023-67019-39169), supporting Lucas Ferreira and Vitor Martins at MSU.