sits (version 0.13.1)

sits_som_cluster: Clustering a set of satellite image time series using SOM

Description

This function uses self-organized maps to find clusters in satellite image time series for quality control of the samples. Calls sits_som_map to generate the som map and sits_som_clean_samples to produce a clean set of samples. The parameters "grid_xdim", "grid_ydim", "rlen", "distance", "alpha", and "iterations" are used by sits_som_map to control how the Kohonen map is generated. The parameters "prior_threshold" and "posterior_threshold" control how the good quality samples are selected, based on the Kohonen map.

Usage

sits_som_cluster(
  data,
  grid_xdim = 10,
  grid_ydim = 10,
  alpha = 1,
  distance = "euclidean",
  rlen = 100,
  prior_threshold = 0.6,
  posterior_threshold = 0.6,
  som_radius = 2
)

Arguments

data

A tibble with samples to be clustered.

grid_xdim

X dimension of the SOM grid (default = 25).

grid_ydim

Y dimension of the SOM grid.

alpha

Starting learning rate, which decreases according to number of iterations.

distance

The similarity measure (distance).

rlen

How many times dataset will be presented to the SOM.

prior_threshold

Threshold of priot probability (frequency of samples assigned to a same SOM neuron)

posterior_threshold

Threshold of posterior probability (influenced by the SOM neighborhood)

som_radius

Radius of neighborhood on the SOM map (controls the size of the neighbourhood)

Value

A sits tibble with an evaluation column indicating if each samples is clean, should be analyzed or should be removed, and with a new column indicating the posterior probability of the sample

References

`kohonen` package (https://CRAN.R-project.org/package=kohonen)

Examples

# NOT RUN {
# Evaluate the quality of the samples using SOM clustering
new_samples <- sits_som_cluster(samples_modis_4bands)
# }