7  Machine learning classification of remote sensing images

7.1 Outline

This chapter describes machine learning methods for classifying individual remote sensing images and image time series. The chapter considers three kinds of algorithms: • Machine learning algorithms that do not explicitly consider the spatial and temporal structure of the time series. These methods include random forests, support vector machine and extreme gradient boosting. • Deep learning methods which consider temporal relations between observed values in a time series. This class of models includes 1D convolutional neural networks and temporal attention-based encoders. • Semantic segmentation methods based on U-net paradigms and multidimensional 2D convolution.