Awesome

segmentInterpreter

A shiny utility to interpret segments of a partitioned time-series. The partitioning is done using the breakpoints function/algorithm from the strucchange R package. The aim of visually interpretting a set of time-series segments is to create a training dataset for automatic classification of the segments into their land dynamics afterwards (using random forest or other classifiers).

Dependencies

R packages spectralResilience, shiny, stringr, RSQLite.

Input data

Input data must be a in a sqlite database. The structure of the table should be as follows. These are extracted individual remote sensing time-series for individual pixels. B1 to B7 correspond to the surface reflectance values of these observations, and sceneID is the Landsat scene identifier of the scene from which the observation was extracted; it's mostly important to retrieve the date of the observation.

Output

The app writes the interpreted segment classes linked to some properties of these segments (mean, phenological amplitude, slope, etc) to a sqlite database.