Land Cover Classification Using Extremely Randomized Trees: A Kernel Perspective

In this research works (2 papers), we evaluate the pros and cons of using tree-based kernels (Random Forest Kernels and Extra-Trees Kernels) in an SVM compared to using the conventional Radial Basis Function (RBF) kernel, standard Random Forest, and standard Extra-Trees classifier. A time series of seven multispectralWorldView-2 images acquired over Sukumba (Mali) are used to illustrate these studies.