Springe direkt zu Inhalt

Methodical Background

In this chapter, we will want to focus on supervised Land Use / Land Cover Classifications (LULC-Classification). In detail, we will have a look at a Classification and Regression Trees-Classifier (CART) and a Random Forest Classifier.

Both of these Algorithms share the trait that they are based on Machine Learning, which means that they use artificial intelligence to generate information and knowledge. This is achieved by first training your classifier in a training phase, ‘feeding’ it examples to learn from, and subsequently generalizing the learned knowledge to either describe or predict a phenomenon of interest.

This method is essential for Remote Sensing scientists, as it enables us to conduct accurate LULC-Classifications of all kinds. In this exercise, we will make use of this to create a Land Use / Land Cover Classification for a research area of our own choice using only the Google Earth Engine and its directly available datasets.

Finally, we will have a look at simple Linear Regressions to analyze trends in time series.