One-Class Classification Of Vegetation Related Changes Via Mutual Ordering Of Normalized Differences

Climate change finds one of its main causes in the happening transition between forest and arid lands of different types as a consequence of wildfires and/or massive deforestation practices. Remote sensing should provide effective and scalable solutions to monitor this dangerous trend. In this paper, a recently developed novel model for one-class classification based on abstract features constructed from normalized difference indices is presented and challenged on detecting deforestation patterns on multispectral images. Results are promising as the performance is nearly optimal, showing that the model comes with good generalization capabilities to deal with vegetation related changes in general.