Multi-temporal land cover maps with a Hidden Markov Model (MTLCHMM)

This is largely a port of S. Parker Abercrombie's code. The main differences between the original code and this library are:

1. MTLCHMM is designed to support any sensor.
2. Parallel processing at the pixel level is handled differently (however, no tests between the two have been conducted).

Reference

Abercrombie, S Parker and Friedl, Mark A (2016) Improving the Consistency of Multitemporal Land Cover Maps Using a Hidden Markov Model. IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, 54(2), 703--713.

Usage

>>> import mtlchmm
>>>
>>> hmm_model = mtlchmm.MTLCHMM(method='forward-backward', 
>>>                             transition_prior=0.1,
>>>                             n_jobs=-1)
>>>
>>> hmm_model.fit_predict(['lc_probas_yr01.tif', 
>>>                        'lc_probas_yr02.tif',
>>>                        'lc_probas_yr03.tif'])

Results from the above example would be written to:

/lc_probas_yr01_hmm.tif
/lc_probas_yr02_hmm.tif
/lc_probas_yr03_hmm.tif

Full example with classification using MpGlue

>>> import mpglue as gl
>>> import mtlchmm
>>>
>>> cl = gl.classification()
>>>
>>> # Load land cover samples.
>>> cl.split_samples('/samples.txt')
>>>
>>> # Train a Random Forest classification model.
>>> cl.construct_model(classifier_info={'classifier': 'RF',
>>>                                     'trees': 1000,
>>>                                     'max_depth': 25})
>>>
>>> lc_probabilities = list()
>>>
>>> # Predict class conditional probabilities and write to file.
>>> for im in ['yr01.tif', 'yr02.tif', 'yr03.tif']:
>>>
>>>     out_probs = '{}_probas.tif'.format(im.split('.')[0])
>>>
>>>     cl.predict(im, 
>>>                out_probs,
>>>                predict_probs=True)
>>>
>>>     # Store output file names.
>>>     lc_probabilities.append(out_probs)
>>>
>>> # Get the class transitional probabilities.
>>> hmm_model = mtlchmm.MTLCHMM(method='forward-backward', 
>>>                             transition_prior=0.1,
>>>                             n_jobs=-1)
>>>
>>> # Fit the HMM model and write 
>>> #   adjusted probabilities to file.
>>> hmm_model.fit_predict(lc_probabilities)

Installation

Dependencies

• NumPy
• MpGlue

Clone the latest version

git clone https://github.com/BU-LCSC/mtlchmm.git
cd mtlchmm
python setup.py install

Updating

cd mtlchmm
git pull origin master
python setup.py install