ABSTRACT
Feature extraction is one of the key steps in Earth observations. Traditional statistics and data science algorithms are limited by their learning capacity and inference mechanism. New artificial intelligence algorithms for machine learning and data mining provide unprecedented opportunities to aid remote sensing image processing in feature extraction. The main objective of this chapter is to focus on the machine learning and data mining algorithms commonly used for feature extraction to process remotely sensed images for supervised, unsupervised, and/or semi-supervised classification. The basic algorithms described in this chapter include:
evolutionary computing-based machine learning techniques such as genetic programming,
artificial neural network-based machine learning techniques such as single-layer forward networks, extreme learning machines, deep believe networks, and convolutional neural networks,
support vector machine, and
optimization-based machine learning techniques, such as genetic algorithm and particle swarm optimization model, in support of hybrid algorithms.
In addition to the theoretical background of these machine learning and data mining techniques, a set of illustrative examples are also included to elevate understanding and promote application potential.
