We will use two possible scenarios in order to show you the possibilities of R when it comes to remote sensing issues:

  1. classification: a common land cover classification of one multispectral Landsat 8 scene
  2. regression: generation of subpixel information based on Landsat 8 data and high resolution reference data

A complete workflow utilizing various R packages is given in the next sections.

Chapter in a Box

In this chapter, the following content awaits you:

Machine Learning Basics
– familiarize yourself with the basic terms of the ML, e.g., unsupervised vs. supervised, linear vs. nonlinear, parametric vs. non-parametric, overfitting vs. underfitting
– learn how to sample polygons in QGIS
– classify a multispectral Landsat 8 scene using a RF and a SVM into several land cover classes
– test the performance of your classifier via a learning curve
– learn how to prepare reference polygons for regressor training
– perform a Support Vector Regression