RStoolbox is an R package providing a wide range of tools for your every-day remote sensing processing needs. The available tool-set covers many aspects from data import, pre-processing, data analysis, image classification and graphical display. RStoolbox builds upon the terra package, which makes it suitable for processing large data-sets even on smaller workstations.
Find out more on the RStoolbox webpage.
The package is available on CRAN and can be installed as usual via
To install the latest version from GitHub you need to have r-base-dev (Linux) or Rtools (Windows) installed. Then run the following lines:
RStoolbox implements a variety of remote sensing methods and workflows. Below are a few examples to get started. Further examples can be found in the documentation of the respective functions.
The example below shows an unsupervised classification workflow based on kmeans clustering:
library(RStoolbox)
# unsupervised classification with 3 classes
uc <- unsuperClass(lsat, nClasses = 3)
# plot result using ggplot
ggR(uc$map, geom_raster = T, forceCat = T) +
scale_fill_manual(values = c("darkgreen", "blue", "sandybrown"))
If training data are available, e.g. labeled polygons, RStoolbox can be used to conduct a supervised classification. The workflow below employs randomForest to train a classification model:
library(RStoolbox)
library(caret)
library(randomForest)
library(ggplot2)
library(terra)
# example: training data from digitized polygons
train <- readRDS(system.file("external/trainingPolygons_lsat.rds", package="RStoolbox"))
# plot input data
ggRGB(lsat, r = 3, g = 2, b=1, stretch = "lin") +
geom_sf(data = train, aes(fill = class)) +
scale_fill_manual(values = c("yellow", "sandybrown", "darkgreen", "blue"))
#> Coordinate system already present. Adding new coordinate system, which will
#> replace the existing one.
# fit random forest (splitting training into 70\% training data, 30\% validation data)
sc <- superClass(lsat, trainData = train, responseCol = "class",
model = "rf", tuneLength = 1, trainPartition = 0.7)
# print model performance and confusion matrix
sc$modelFit
#> [[1]]
#> TrainAccuracy TrainKappa method
#> 1 0.9992293 0.9988032 rf
#>
#> [[2]]
#> Cross-Validated (5 fold) Confusion Matrix
#>
#> (entries are average cell counts across resamples)
#>
#> Reference
#> Prediction cleared fallen_dry forest water
#> cleared 141.6 0.0 0.0 0.0
#> fallen_dry 0.0 22.0 0.0 0.0
#> forest 0.4 0.0 255.0 0.0
#> water 0.0 0.0 0.0 99.4
#>
#> Accuracy (average) : 0.9992
# plotting: convert class IDs to class labels (factorize) and plot
r <- as.factor(sc$map)
levels(r) <- data.frame(ID = 1:4, class_supervised = levels(train$class))
ggR(r, geom_raster = T, forceCat = T) + scale_fill_manual(values = c("yellow", "sandybrown", "darkgreen", "blue"))
Created on 2024-04-19 with reprex v2.1.0
RStoolbox offers spectral unmixing by implementing the Multiple Endmember Spectral Mixture Analysis (MESMA) approach for estimating fractions of spectral classes, such as spectra of surfaces or materials, on a sub-pixel scale. The following workflow shows a simple Spectral Mixture Analysis (SMA) with single endmembers per class, extracted from the lsat example image by cell id:
library(RStoolbox)
library(terra)
# to perform a SMA, use a single endmember per class, row by row:
em <- data.frame(lsat[c(5294, 47916)])
rownames(em) <- c("forest", "water")
# umix the lsat image
probs <- mesma(img = lsat, em = em)
plot(probs)
Instead, one can define multiple endmembers per class to conduct a Multiple Endmember Spectral Mixture Analysis (MESMA):
library(RStoolbox)
library(terra)
# to perform a MESMA, use multiple endmembers per class, differntiating them
# by a column named 'class':
em <- rbind(
data.frame(lsat[c(4155, 17018, 53134, 69487, 83704)], class = "forest"),
data.frame(lsat[c(22742, 25946, 38617, 59632, 67313)], class = "water")
)
# unmix the lsat image
probs <- mesma(img = lsat, em = em)
plot(probs)
# MESMA can also be performed on more than two endmember classes:
em <- rbind(
data.frame(lsat[c(4155, 17018, 53134, 69487, 83704)], class = "forest"),
data.frame(lsat[c(22742, 25946, 38617, 59632, 67313)], class = "water"),
data.frame(lsat[c(4330, 1762, 1278, 1357, 17414)], class = "shortgrown")
)
# unmix the lsat image
probs <- mesma(img = lsat, em = em)
plot(probs)
RStoolbox comes with a suite of pre-processing functions, including cloudMask to identify clouds in optical satellite imagery:
library(ggplot2)
# lsat example scene, with two tiny clouds in the east
ggRGB(lsat, stretch = "lin")
# calculate cloud index
cldmsk <- cloudMask(lsat, blue = 1, tir = 6)
ggR(cldmsk, 2, geom_raster = TRUE) 
# mask by threshold, region-growing around the core cloud pixels
cldmsk_final <- cloudMask(cldmsk, threshold = 0.1, buffer = 5)
## plot cloudmask
ggRGB(lsat, stretch = "lin") +
ggR(cldmsk_final[[1]], ggLayer = TRUE, forceCat = TRUE, geom_raster = TRUE) +
scale_fill_manual(values = c("red"), na.value = NA)
#> Warning: Removed 88752 rows containing missing values or values outside the scale range
#> (`geom_raster()`).
With radCor, users can compute radiometric and simple atmospheric corrections (based on dark object substraction):
library(terra)
# import Landsat meta data
mtlFile <- system.file("external/landsat/LT52240631988227CUB02_MTL.txt", package="RStoolbox")
metaData <- readMeta(mtlFile)
lsat_t <- stackMeta(mtlFile)
# convert DN to top of atmosphere reflectance and brightness temperature
lsat_ref <- radCor(lsat_t, metaData = metaData, method = "apref")
# correct DN to at-surface-reflecatance with DOS (Chavez decay model)
lsat_sref <- radCor(lsat_t, metaData = metaData)
# correct DN to at-surface-reflecatance with simple DOS and automatic haze estimation
hazeDN <- estimateHaze(lsat_t, hazeBands = 1:4, darkProp = 0.01, plot = FALSE)
lsat_sref <- radCor(lsat_t, metaData = metaData, method = "sdos",
hazeValues = hazeDN, hazeBands = 1:4)
# plot result
ggRGB(lsat_sref, r = 3, g = 2, b = 1, stretch = "lin")
Created on 2024-04-19 with reprex v2.1.0