| Type: | Package |
| Version: | 0.2.1 |
| Title: | Bayesian Smoothing of Remote Sensing Image Classification |
| Maintainer: | Gilberto Camara <gilberto.camara.inpe@gmail.com> |
| Description: | A Bayesian smoothing method for post-processing of remote sensing image classification which refines the labelling in a classified image in order to enhance its classification accuracy. Combines pixel-based classification methods with a spatial post-processing method to remove outliers and misclassified pixels. |
| Encoding: | UTF-8 |
| Language: | en-US |
| Depends: | R (≥ 4.0.0) |
| URL: | https://github.com/e-sensing/bayesEO/ |
| BugReports: | https://github.com/e-sensing/bayesEO/issues |
| License: | GPL-3 |
| ByteCompile: | true |
| Imports: | dplyr, ggplot2, grDevices, purrr, Rcpp, stars, stats, terra, tibble, tidyr, tmap, yaml |
| Suggests: | RcppArmadillo, testthat |
| LinkingTo: | Rcpp, RcppArmadillo |
| RoxygenNote: | 7.3.1 |
| Collate: | 'api_check.R' 'bayes_colors.R' 'bayes_label.R' 'bayes_plot.R' 'bayes_read.R' 'bayes_smooth.R' 'bayes_utils.R' 'bayes_variance.R' 'RcppExports.R' 'zzz.R' |
| NeedsCompilation: | yes |
| Packaged: | 2024-06-02 02:25:45 UTC; gilberto |
| Author: | Gilberto Camara [aut, cre], Renato Assuncao [aut], Rolf Simoes [aut], Felipe Souza [aut] |
| Repository: | CRAN |
| Date/Publication: | 2024-06-04 09:44:41 UTC |
Compute the 2-D Gaussian kernel
Description
Compute the 2-D Gaussian kernel
Usage
.smooth_gauss_kernel(window_size, sigma)
Arguments
window_size |
Size of the neighbourhood. |
sigma |
Standard deviation of the spatial Gaussian kernel |
Value
returns a squared matrix filled with Gaussian function
Function to retrieve bayesEO color table
Description
Returns a color table
Usage
bayes_colors()
Value
A tibble with color names and values
Author(s)
Gilberto Camara, gilberto.camara@inpe.br
Function to show colors in SITS
Description
Shows the default SITS colors
Usage
bayes_colors_show()
Value
no return, called for side effects
Author(s)
Gilberto Camara, gilberto.camara@inpe.br
Label probability images to create categorical maps
Description
Takes a classified image with probabilities, and labels the image with the pixel of higher probability
Usage
bayes_label(x)
Arguments
x |
SpatRaster object with probabilities images |
Value
A SpatRaster object
Author(s)
Gilberto Camara, gilberto.camara@inpe.br
Examples
if (bayes_run_examples()) {
# select a file with probability values
data_dir <- system.file("/extdata/probs/", package = "bayesEO")
file <- list.files(data_dir)
# create a SpatRaster object from the file
probs_file <- paste0(data_dir, "/", file)
# provide the labels
labels <- c("Water", "ClearCut_Burn", "ClearCut_Soil",
"ClearCut_Veg", "Forest", "Wetland")
# read the probs file
probs <- bayes_read_probs(probs_file, labels)
# produce a labelled map
map <- bayes_label(probs)
# plot the labelled map
bayes_plot_map(map)
}
Plot histogram
Description
Plot histogram
Usage
bayes_plot_hist(x, scale = 1, quantile = NULL, sample_size = 15000)
Arguments
x |
SpatRaster to be plotted. |
scale |
Scale factor for SpatRaster |
quantile |
Threshold of values that will be plotted |
sample_size |
Number of samples to extract values |
Value
A plot object
Author(s)
Gilberto Camara, gilberto.camara@inpe.br
Examples
if (bayes_run_examples()) {
# get the probability file
data_dir <- system.file("/extdata/probs/", package = "bayesEO")
file <- list.files(data_dir)
# read the probability file into a SpatRaster
x <- terra::rast(paste0(data_dir, "/", file))
# include the labels
labels <- c("Water", "ClearCut_Burn", "ClearCut_Soil",
"ClearCut_Veg", "Forest", "Wetland")
# associate the labels to the names of the SpatRaster
names(x) <- labels
# calculate the variance
v <- bayes_variance(x)
# Plot the variance histogram
bayes_hist(v, quantile = 0.75)
}
Plot labelled map
Description
Plot labelled map
Usage
bayes_plot_map(
x,
legend = NULL,
palette = "Spectral",
xmin = NULL,
xmax = NULL,
ymin = NULL,
ymax = NULL,
tmap_graticules_labels_size = 0.6,
tmap_legend_title_size = 0.7,
tmap_legend_text_size = 0.7,
tmap_legend_bg_color = "white",
tmap_legend_bg_alpha = 0.5,
tmap_max_cells = 1e+06
)
Arguments
x |
SpatRaster to be plotted. |
legend |
Named vector that associates labels to colors. |
palette |
A sequential RColorBrewer palette |
xmin |
Subset to be shown (xmin) |
xmax |
Subset to be shown (xmax) |
ymin |
Subset to be shown (ymin) |
ymax |
Subset to be shown (ymax) |
tmap_graticules_labels_size |
Size of graticules labels (default: 0.7) |
tmap_legend_title_size |
Size of legend title (default: 1.5) |
tmap_legend_text_size |
Size of legend text (default: 1.2) |
tmap_legend_bg_color |
Color of legend backgound (default: "white") |
tmap_legend_bg_alpha |
Transparency of legend background (default: 0.5) |
tmap_max_cells |
Maximum number of cells for tmap (default = 1e+06) |
Value
A plot object
Author(s)
Gilberto Camara gilberto.camara@inpe.br
Examples
if (bayes_run_examples()) {
# Define location of a probability file
data_dir <- system.file("/extdata/probs",
package = "bayesEO")
# list the file
file <- list.files(data_dir)
# build the full path
probs_file <- paste0(data_dir, "/", file)
# define labels
labels <- c("Water", "ClearCut_Burn", "ClearCut_Soil",
"ClearCut_Veg", "Forest", "Wetland")
probs_image <- bayes_read_probs(probs_file, labels)
# Label the probs image
y <- bayes_label(x)
# produce a map of the labelled image
bayes_plot_map(y)
}
Plot probability maps
Description
Plot probability maps
Usage
bayes_plot_probs(
x,
scale = 1e-04,
labels = NULL,
palette = "YlGnBu",
tmap_scale = 1
)
Arguments
x |
SpatRaster to be plotted. |
scale |
Scaling factor to apply to the data |
labels |
Labels to be plotted |
palette |
An RColorBrewer palette |
tmap_scale |
Global scale parameter for map (default: 1.0) |
Value
A plot object
Author(s)
Gilberto Camara, gilberto.camara@inpe.br
Examples
if (bayes_run_examples()) {
# get the probability file
data_dir <- system.file("/extdata/probs/", package = "bayesEO")
file <- list.files(data_dir)
# build the full path
probs_file <- paste0(data_dir, "/", file)
# include the labels
labels <- c("Water", "ClearCut_Burn", "ClearCut_Soil",
"ClearCut_Veg", "Forest", "Wetland")
# associate the labels to the names of the SpatRaster
probs <- bayes_read_probs(probs_file, labels)
# Plot the probability image
bayes_plot_probs(probs,
scale = 0.0001,
tmap_scale = 1.0)
}
Plot RGB data cubes
Description
Plot RGB raster cube
Usage
bayes_plot_rgb(
image,
red,
green,
blue,
xmin = NULL,
xmax = NULL,
ymin = NULL,
ymax = NULL
)
Arguments
image |
Object of class SpatRaster. |
red |
Band for red color. |
green |
Band for green color. |
blue |
Band for blue color. |
xmin |
Subset to be shown (xmin) |
xmax |
Subset to be shown (xmax) |
ymin |
Subset to be shown (ymin) |
ymax |
Subset to be shown (ymax) |
Value
A plot object with an RGB image
Author(s)
Gilberto Camara, gilberto.camara@inpe.br
Examples
if (bayes_run_examples()) {
# Define location of a RGB files
rgb_dir <- system.file("/extdata/rgb", package = "bayesEO")
# list the file
files <- list.files(rgb_dir)
# build the full path
image_files <- paste0(rgb_dir, "/", files)
rgb_image <- bayes_read_image(image_files)
bayes_plot_rgb(rgb_image, red = "B11", green = "B8A", blue = "B03")
}
Plot variance maps
Description
Plot variance maps
Usage
bayes_plot_var(
x,
labels = NULL,
quantile = 0.75,
n = 15,
style = "equal",
palette = "YlGnBu",
tmap_scale = 1
)
Arguments
x |
SpatRaster to be plotted. |
labels |
Labels to be plotted |
quantile |
Thereshold of values to be plotted |
n |
Preferred number of classes |
style |
Method to process the color scale |
palette |
An RColorBrewer palette |
tmap_scale |
Global scale parameter for map (default: 1.5) |
Value
A plot object
Author(s)
Gilberto Camara, gilberto.camara@inpe.br
Examples
if (bayes_run_examples()) {
# get the probability file
data_dir <- system.file("/extdata/probs/", package = "bayesEO")
file <- list.files(data_dir)
# build the full path
probs_file <- paste0(data_dir, "/", file)
# include the labels
labels <- c("Water", "ClearCut_Burn", "ClearCut_Soil",
"ClearCut_Veg", "Forest", "Wetland")
# associate the labels to the names of the SpatRaster
probs <- bayes_read_probs(probs_file, labels)
# calculate the variance
var <- bayes_variance(probs)
# Plot the variance image
bayes_plot_var(var,
n = 15,
style = "order",
quantile = 0.75,
palette = "YlGn",
labels = c("Forest", "ClearCut_Veg"))
}
Read probability maps
Description
Read probability maps
Usage
bayes_read_image(files)
Arguments
files |
Full path to raster files |
Value
A SpatRaster object
Author(s)
Gilberto Camara, gilberto.camara@inpe.br
Examples
if (bayes_run_examples()) {
# Define location of a probability file
data_dir <- system.file("/extdata/rgb", package = "bayesEO")
# list the file
files <- list.files(data_dir)
# build the full path
image_files <- paste0(data_dir, "/", files)
rgb_image <- bayes_read_image(image_files)
}
Read probability maps
Description
Read probability maps
Usage
bayes_read_probs(probs_file, labels)
Arguments
probs_file |
Full path to raster multi-band file containing probability matrices |
labels |
Labels to be assigned to the bands |
Value
A SpatRaster object
Author(s)
Gilberto Camara, gilberto.camara@inpe.br
Examples
if (bayes_run_examples()) {
# Define location of a probability file
data_dir <- system.file("/extdata/probs", package = "bayesEO")
# list the file
file <- list.files(data_dir)
# build the full path
probs_file <- paste0(data_dir, "/", file)
# define labels
labels <- c("Water", "ClearCut_Burn", "ClearCut_Soil",
"ClearCut_Veg", "Forest", "Wetland")
probs_image <- bayes_read_probs(probs_file, labels)
}
Informs if examples should run
Description
This function informs if examples should run. To run the examples, set "BAYES_RUN_EXAMPLES" environment variable to "YES" using Sys.setenv("BAYES_RUN_EXAMPLES" = "YES") To come back to the default behaviour, please unset the enviroment variable Sys.unsetenv("BAYES_RUN_EXAMPLES")
Usage
bayes_run_examples()
Value
A logical value
Informs if tests should run
Description
This function informs if tests should run. To run the examples, set "BAYES_RUN_TESTS" environment variable to "YES" using Sys.setenv("BAYES_RUN_TESTS" = "YES") To come back to the default behaviour, please unset the enviroment variable Sys.unsetenv("BAYES_RUN_TESTS")
Usage
bayes_run_tests()
Value
TRUE/FALSE
Smooth probability images
Description
Takes a classified image with probabilities, and reduces outliers and smoothens probability according to Bayesian statistics
Usage
bayes_smooth(x, window_size = 7, neigh_fraction = 0.5, smoothness = 10)
Arguments
x |
SpatRaster object with probabilities images |
window_size |
Size of the neighborhood. |
neigh_fraction |
Fraction of neighbors with high probabilities to be used in Bayesian inference. |
smoothness |
Estimated variance of logit of class probabilities (Bayesian smoothing parameter). It can be either a vector or a scalar. |
Value
A SpatRaster object
Author(s)
Gilberto Camara, gilberto.camara@inpe.br
Examples
if (bayes_run_examples()) {
# select a file with probability values
data_dir <- system.file("/extdata/probs/", package = "bayesEO")
file <- list.files(data_dir)
# create a full path for the file
probs_file <- paste0(data_dir, "/", file)
# provide the labels
labels <- c("Water", "ClearCut_Burn", "ClearCut_Soil",
"ClearCut_Veg", "Forest", "Wetland")
# read the probs file
probs <- bayes_read_probs(probs_file, labels)
# smooth the probability image
probs_smooth <- bayes_smooth(probs,
window_size = 7,
smoothness = 20
)
# plot the probability image
bayes_plot_probs(probs_smooth)
}
Summary of categorical maps
Description
Takes a classified image with probabilities, and labels the image with the pixel of higher probability
Usage
bayes_summary(x, scale = 1, sample_size = 15000)
Arguments
x |
SpatRaster categorical object |
scale |
Scale to apply to data |
sample_size |
Sample size |
Value
A tibble with information
Author(s)
Gilberto Camara, gilberto.camara@inpe.br
Examples
if (bayes_run_examples()) {
# select a file with probability values
data_dir <- system.file("/extdata/probs/", package = "bayesEO")
file <- list.files(data_dir)
# create a SpatRaster object from the file
probs_file <- paste0(data_dir, "/", file)
# provide the labels
labels <- c("Water", "ClearCut_Burn", "ClearCut_Soil",
"ClearCut_Veg", "Forest", "Wetland")
# read the probs file
probs <- bayes_read_probs(probs_file, labels)
# produce a labelled map
map <- bayes_label(probs)
# plot the labelled map
bayes_summary(map)
}
Calculate the variance of a probability cube
Description
Takes a probability cube and estimate the local variance of the logit of the probability, to support the choice of parameters for Bayesian smoothing.
Usage
bayes_variance(x, window_size = 9, neigh_fraction = 0.5)
Arguments
x |
SpatRaster object containing probabilities. |
window_size |
Size of the neighborhood. |
neigh_fraction |
Fraction of neighbors with highest probability to be used in Bayesian inference. |
Value
A variance SpatRaster object.
Author(s)
Gilberto Camara, gilberto.camara@inpe.br
Rolf Simoes, rolf.simoes@inpe.br
Examples
if (bayes_run_examples()) {
# select a file with probability values
data_dir <- system.file("/extdata/probs/", package = "bayesEO")
file <- list.files(data_dir)
# create a SpatRaster object from the file
x <- terra::rast(paste0(data_dir, "/", file))
# provide the labels
labels <- c("Water", "ClearCut_Burn", "ClearCut_Soil",
"ClearCut_Veg", "Forest", "Wetland")
# name the layers in the SpatRaster with the labels
names(x) <- labels
# calculate the variance
v <- bayes_variance(x)
# plot the variance
bayes_plot_var(v, quantile = 0.75)
}
Smooth probability images with Gaussian filter
Description
Takes a classified image with probabilities, and reduces outliers and smoothens probability according to a Gaussian filter
Usage
bilateral_smooth(x, window_size = 5, sigma = 8, tau = 0.1)
Arguments
x |
SpatRaster object with probabilities images |
window_size |
Size of the neighborhood. |
sigma |
Standard deviation of the spatial Gaussian kernel |
tau |
Standard deviation of the class probs value |
Value
A SpatRaster object
Author(s)
Gilberto Camara, gilberto.camara@inpe.br
Examples
if (bayes_run_examples()) {
# select a file with probability values
data_dir <- system.file("/extdata/probs/", package = "bayesEO")
file <- list.files(data_dir)
# create a full path for the file
probs_file <- paste0(data_dir, "/", file)
# provide the labels
labels <- c("Water", "ClearCut_Burn", "ClearCut_Soil",
"ClearCut_Veg", "Forest", "Wetland")
# read the probs file
probs <- bayes_read(probs_file, labels)
# smooth the probability image
bilat <- bilateral_smooth(probs,
window_size = 5,
sigma = 8,
tau = 0.1
)
# plot the probability image
bayes_plot(bilat, scale = 0.0001)
}
Smooth probability images with Gaussian filter
Description
Takes a classified image with probabilities, and reduces outliers and smoothens probability according to a Gaussian filter
Usage
gaussian_smooth(x, window_size = 5, sigma = 1)
Arguments
x |
SpatRaster object with probabilities images |
window_size |
Size of the neighborhood. |
sigma |
Standard deviation of the spatial Gaussian kernel |
Value
A SpatRaster object
Author(s)
Gilberto Camara, gilberto.camara@inpe.br
Examples
if (bayes_run_examples()) {
# select a file with probability values
data_dir <- system.file("/extdata/probs/", package = "bayesEO")
file <- list.files(data_dir)
# create a full path for the file
probs_file <- paste0(data_dir, "/", file)
# provide the labels
labels <- c("Water", "ClearCut_Burn", "ClearCut_Soil",
"ClearCut_Veg", "Forest", "Wetland")
# read the probs file
probs <- bayes_read(probs_file, labels)
# smooth the probability image
gauss <- gaussian_smooth(probs,
window_size = 5,
sigma = 1
)
# plot the probability image
bayes_plot_probs(gauss)
}