jgrss / geowombat

GeoWombat: Utilities for geospatial data

Like a wombat, GeoWombat has a simple interface (for raster I/O) with a strong backend (for data processing at scale).

Common Remote Sensing Uses

• Simple read/write for a variety of sensors, including:
  • Sentinel 2
  • Landsat 5-8
  • PlanetScope
  • Others
• Image mosaicking
• On-the-fly image transformations (reprojection)
• Point / polygon raster sampling, extraction
• Time series analysis
• Band math (NDVI, Tasseled cap, EVI etc)
• Image classification and regression
• Radiometry (BRDF normalization)
• Distributed processing

Basic usage - Sentinel & Landsat

>>> import geowombat as gw

Use a context manager and Xarray plotting to analyze processing chains

>>> # Define satellite sensors (here, Landsat 7)
>>> with gw.config.update(sensor='l7'):
>>>
>>>     # Open images as Xarray DataArrays
>>>     with gw.open('LT05_L1TP_227083_20110123_20161011_01_T1.tif') as src:
>>>
>>>         # Apply calculations using Xarray and Dask
>>>         results = src.sel(band=['blue', 'green', 'red']).mean(dim='band')
>>>
>>>         # Check results by computing the task and plotting
>>>         results.gw.imshow()

Use a context manager to pass sensor information to geowombat methods

>>> # Set the sensor as Sentinel 2
>>> with gw.config.update(sensor='s2'):
>>>
>>>     # Open a Sentinel 2 image
>>>     with gw.open('L1C_T20HPH_A002352_20151204T141125_MTD.tif') as src:
>>>
>>>         # Use built-in normalization methods, such as the NDVI
>>>         ndvi = src.gw.ndvi(scale_factor=0.0001)
>>>
>>>         # Check results by computing the task and plotting
>>>         ndvi.gw.imshow()

Computation scales easily over large datasets with minimal changes to the code.

>>> # Set a reference image to align to
>>> with gw.config.update(ref_image='ref_image.tif'):
>>>
>>>     # Open images as Xarray DataArrays
>>>     with gw.open('image_a.tif') as srca, gw.open('image_b.tif') as srcb:
>>>
>>>         # The size of srca, srcb, and results are determined by the configuration context
>>>         results = srca.sel(band=1) * srcb.sel(band=[1, 2, 3]).mean(dim='band')
>>>
>>>         # Initiate computation by writing the results to file. 
>>>         # Mix process and thread workers to execute the task in parallel. 
>>>         results.gw.to_raster('output.tif', 
>>>                              n_workers=4, 
>>>                              n_threads=4,
>>>                              compress='lzw')

Installation

GeoWombat is not on PyPi, but it can be installed with pip. Most dependencies will be automatically installed. However, the exception is GDAL, which needs to be installed prior to executing the pip command below. See the installation documentation for details on installing GDAL. With GDAL installed, GeoWombat can be installed like:

pip install git+https://github.com/jgrss/geowombat

Alternatively, use git to clone and build the repository like:

git clone https://github.com/jgrss/geowombat.git
cd geowombat/
python setup.py build && pip install --upgrade . && rm -rf build/

opencv requires libGL.so.1. In Ubuntu, this can be obtained by running:

apt-get update && apt-get install libgl1 -y

Update

To update GeoWombat:

pip install --upgrade git+https://github.com/jgrss/geowombat

Or, to update a cloned repository:

cd geowombat/
git pull origin master
python setup.py build && pip install --upgrade . && rm -rf build/

(Optional) Install into a virtual environment on Linux

With virtualenv

# Create a Python 3.7 virtual environment named gwenv
virtualenv -p python=3.7 gwenv

# Activate the virtual environment
source gwenv/bin/activate

# Install the libspatialindex and GDAL libraries
sudo apt install libspatialindex-dev libgdal-dev

# Install GeoWombat with all extra libraries
pip install git+https://github.com/jgrss/geowombat.git#egg=project[all]

With conda

# Install the libspatialindex and GDAL libraries
sudo apt install libspatialindex-dev libgdal-dev

# Create a Python 3.7 virtual environment named gwenv
conda create -n gwenv python=3.7 cython numpy libspatialindex gdal -c conda-forge

# Activate the virtual environment
conda activate gwenv

# Install GeoWombat with all extra libraries
pip install git+https://github.com/jgrss/geowombat.git

Universal Install Via Docker

If you are having trouble installing geowombat, the surest way to get it up and running is with Docker containers. Please follow the instructions on pygis.io.

Development: Adding packages

Install pip-tools. Add packages to requirements.in and running the following to compile requirements.txt:

pip-compile requirements.in

Documentation

For more details, see https://geowombat.readthedocs.io.

Learning

If you are new to geospatial programming in Python please refer to pygis.io