unmixing

Interactive tools for spectral mixture analysis of multispectral raster data in Python

https://github.com/arthur-e/unmixing

Keywords

Scientific Fields

Repository

Interactive tools for spectral mixture analysis of multispectral raster data in Python

README.rst

.. image:: https://zenodo.org/badge/DOI/10.5281/zenodo.3585979.svg
   :target: https://doi.org/10.5281/zenodo.3585979

==========================
Interactive Unmixing Tools
==========================

Overview
========

This is a library of interactive tools and functions for performing linear spectral mixture analysis (LSMA) and spatially adaptive spectral mixture analysis (SASMA).
It supports parallel fully constrained least-squares (FCLS) mixture analysis over multiple processes, allowing for very efficient mapping of endmember abundances, both in the spatially adaptive approach and in regular LSMA.
In detail, the `unmixing` module includes tools and functions to support:

- Stacking raster bands from any GDAL file format, including HDF;
- Applying a mask or a combination of masks to a raster;
- Generating raster masks from saturated pixels or using density slicing (e.g., to mask water pixels);
- Applying CFMask output and other quality assurance layers from the USGS;
- Band-wise compositing of reflectance (or fraction/ abundance images);
- Tasseled cap transformation and other indices including NDVI and RNDSI;
- Dimensionality reduction through the Minimum Noise Fraction (MNF);
- Radiometric rectification of a raster time series;
- Visualizing the mixing space of a moderate resolution raster;
- Interactive selection of pixels in the mixing space, which are then captured in a KML file;
- Endmember induction and plotting of endmembers for linear spectral mixture analysis (LSMA);
- Fully constrained least squares (FCLS) unmixing;
- Spatial interpolation of endmembers for spatially adaptive spectral mixture analysis (SASMA);
- Learning of endmember candidates for SASMA using classification and regression trees (CART);
- Validation of unmixing through a forward model of reflectance;

**For an overview and tutorial on how to use this library, check out the iPython Notebook(s) in the** ``docs/`` **folder:**

- 
- 

Installation and Setup
======================

Because this is a scientific library, there are complex dependencies that may be difficult to install.
For GNU/Linux systems, particularly Ubuntu, look at ``install.sh`` for a guide on installing the system dependencies required for the Python dependencies.

At this time, installation is intended for development purposes only.
As such, ``unmixing`` should be installed in "editable" mode using ``pip``; see .
From the ``unmixing`` directory, where ``setup.py`` is found::

    $ pip install -e .

Dependencies
------------

* ``numpy``
* ``scipy``
* ``matplotlib``
* ``pysptools``
* ``GDAL``
* ``pykml``
* ``scikit-learn`` (For SASMA)

Use System Dependencies
-----------------------

Many of the core packages, particularly NumPy and SciPy, have wide adoption and use.
They also take a lot of time and clock cycles to compile into a virtual environment.
Consequently, it is recommended that these libraries be installed globally (system-wide).

Other packages should be installed only in the virtual environment.
To use both system-wide and local packages within a virtual environment, the virtual environment must be set up with the ``--system-site-packages`` option::

    sudo apt-get install python3-numpy python3-scipy python3-matplotlib python3-zmq
    virtualenv -p /usr/bin/python3.5 --system-site-packages 

The local packages can be installed within the virtual environment via ``pip``::

    source /my/virtualenv/bin/activate
    pip install -r REQUIREMENTS


Documentation
=============

Python Dependencies
-------------------

Some resources on the dependencies can be found:

* ``matplotlib`` 
* PySptools 
* PySptools on PyPI 

Third-Party Tools
-----------------

Documentation on related (but not required) third-party tools can be found:

* Fmask 
* CFmask  (A C version of Fmask)
* LEDAPS 

Foundations
===========

Hyperspectral Imaging (HSI) Cubes
---------------------------------

Many of the tools in this library are designed to work with HSI cubes.
While the raster arrays read in from GDAL are ``p x m x n`` arrays, HSI cubes are ``n x m x p`` arrays, where ``m`` is the row index, `n` is the column index, and ``p`` is the spectral band number.
It is important to note that the row index `n` corresponds to the latitude while the column index `m` corresponds to the longitude.
Thus, the coordinates ``(n, m)`` can be converted directly to a longitude-latitude pair.

Owner

Citation (CITATION.md)

# How to Cite

[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.3585979.svg)](https://doi.org/10.5281/zenodo.3585979)

To cite "unmixing" in a scholarly article, please use the DOI `10.5281/zenodo.3585979`, preferably as part of a full citation:

```
Endsley, K. A. 2019. "The unmixing library: Interactive tools for spectral mixture analysis of multispectral raster data in Python v0.2.4.dev." Zenodo. https://zenodo.org/record/3585979
```

GitHub Events

Committers

Last synced: almost 2 years ago

Issues and Pull Requests

Last synced: almost 2 years ago