PyCCD - Python Continuous Change Detection

pyccd creates change segments from Landsat timeseries data

Using PyCCD

>>> import ccd
>>> results = ccd.detect(dates, blues, greens, reds, nirs, swir1s, swir2s, thermals, qas, prev_results)
>>>
>>> type(results)
<class 'dict'>
>>>
>>> results
{algorithm: 'pyccd:x.x.x',
 processing_mask: [bool, bool, ...],
 snow_prob: float,
 water_prob: float,
 cloud_prob: float,
 change_models: [
     {start_day: int,
      end_day: int,
      break_day: int,
      observation_count: int,
      change_probability: float,
      curve_qa: int,
      blue:      {magnitude: float,
                  rmse: float,
                  coefficients: (float, float, ...),
                  intercept: float},
      green:     {magnitude: float,
                  rmse: float,
                  coefficients: (float, float, ...),
                  intercept: float},
      red:       {magnitude: float,
                  rmse: float,
                  coefficients: (float, float, ...),
                  intercept: float},
      nir:       {magnitude: float,
                  rmse: float,
                  coefficients: (float, float, ...),
                  intercept: float},
      swir1:     {magnitude: float,
                  rmse: float,
                  coefficients: (float, float, ...),
                  intercept: float},
      swir2:     {magnitude: float,
                  rmse: float,
                  coefficients: (float, float, ...),
                  intercept: float},
      thermal:   {magnitude: float,
                  rmse: float,
                  coefficients: (float, float, ...),
                  intercept: float}}
                 ]
}

Default processing parameters can be overridden using a dictionary (see parameters.yaml for valid keys):

>>> import ccd
>>> params = {'QA_BITPACKED': False,
              'QA_FILL': 255,
              'QA_CLEAR': 0,
              'QA_WATER': 1,
              'QA_SHADOW': 2,
              'QA_SNOW': 3,
              'QA_CLOUD': 4}
>>> results = ccd.detect(dates, blues, greens, reds, nirs, swir1s, swir2s, thermals, qas, params=params)

Installing

System requirements (Ubuntu)

• python3-dev
• gfortran
• libopenblas-dev
• liblapack-dev
• graphviz
• python-virtualenv

System requirements (Centos)

• python3-devel
• gfortran
• blas-dev
• lapack-dev
• graphviz
• python-virtualenv

It's highly recommended to do all your development & testing in a virtual environment.

user@dev:/home/user/$ mkdir pyccd
user@dev:/home/user/$ cd pyccd
user@dev:/home/user/pyccd$ virtualenv -p python3 .venv
user@dev:/home/user/pyccd$ . .venv/bin/activate
(.venv) user@dev:/home/user/pyccd$

Install

$ pip install -e .[test,dev,docs,deploy,profile]

Testing

$ pytest
$ pytest --profile
$ pytest --profile-svg

# pytest-watch
$ ptw

Make targets

$ make build
$ make tests
$ make docs
$ make deploy
$ make profile

Profiling

Decorate the function to be profiled with @profile and run make profile. Remove decorations before committing code.

Contributing

Before committing to this repository, run the following command.

git config --local commit.template .gitmessage

This will add the LCMAP commit template to git commit.

jira/lcmap-xxxx: Title
Description

Contributions are most welcome.

1. Open an issue and discuss the change.
2. Branch from major version and name it after the issue
   • jira/lcmap-xxxx
3. Write automated tests for your changes and make sure all tests pass.
4. Update documentation in project.
5. Submit pull request to the appropriate branch (e.g. 1,x)

Versions

PyCCD previously followed MAJOR.MINOR.PATCH.LABEL semantic versioning but has changed to date based semantic versioning, thus: YYYY.MM.DD[.HH.MM.SS][-label].

PyCCD's version is defined by the ccd/version.py/__version__ attribute.

References

• PyCCD Digital Object Identifier 10.5066/P90V8IIX
• Test Data
• Landsat Band Specifications
• Landsat 8 Surface Reflectance Specs
• Landsat 4-7 Surface Reflectance Specs
• Landsat Analysis Ready Data
• LCMAP CCDC Collection 1.0 Algorithm Description Document