Isolation forests to evaluate class separability and the representativeness of training and validation areas in land cover classification

Alonso-Sarria, Francisco; Valdivieso Ros, Carmen; Gomariz Castillo, Francisco

Publication:
Isolation forests to evaluate class separability and the representativeness of training and validation areas in land cover classification

dc.contributor.author	Alonso-Sarria, Francisco
dc.contributor.author	Valdivieso Ros, Carmen
dc.contributor.author	Gomariz Castillo, Francisco
dc.contributor.department	Geografía
dc.date.accessioned	2024-12-29T12:20:59Z
dc.date.available	2024-12-29T12:20:59Z
dc.date.issued	2019-12-13
dc.description	© 2019 by the authors. This manuscript version is made available under the CC-BY 4.0 license http://creativecommons.org/licenses/by/4.0/ This document is the Published Manuscript version of a Published Work that appeared in final form in Remote Sensing. To access the final edited and published work see https://doi.org/10.3390/rs11243000
dc.description.abstract	Supervised land cover classification from remote sensing imagery is based on gathering a set of training areas to characterise each of the classes and to train a predictive model that is then used to predict land cover in the rest of the image. This procedure relies mainly on the assumptions of statistical separability of the classes and the representativeness of the training areas. This paper uses isolation forests, a type of random tree ensembles, to analyse both assumptions and to easily correct lack of representativeness by digitising new training areas where needed to improve the classification of a Landsat-8 set of images with Random Forest. The results show that the improved set of training areas after the isolation forest analysis is more representative of the whole image and increases classification accuracy. Besides, the distribution of isolation values can be useful to estimate class separability. A class separability parameter that summarises such distributions is proposed. This parameter is more correlated to omission and commission errors than other separability measures such as the Jeffries–Matusita distance.	es
dc.format	application/pdf	es
dc.format.extent	21	es
dc.identifier.citation	Remote Sensing, 2019, Vol. 11 (24) : 3000
dc.identifier.doi	https://doi.org/10.3390/rs11243000
dc.identifier.issn	Electronic: 2072-4292
dc.identifier.uri	http://hdl.handle.net/10201/147874
dc.language	eng	es
dc.publisher	MDPI	es
dc.relation	This research was funded by the Spanish Ministerio de Economía, Industria y Competitividad (MINECO), la Agencia Estatal de Investigación (AEI) and Fondo Europeo de Desarrollo Regional (FEDER). Project CGL2017-84625-C2-2-R	es
dc.relation.publisherversion	https://www.mdpi.com/2072-4292/11/24/3000	es
dc.rights	info:eu-repo/semantics/openAccess	es
dc.rights	Atribución 4.0 Internacional	*
dc.rights.uri	http://creativecommons.org/licenses/by/4.0/	*
dc.subject	Training area representativeness	es
dc.subject	Class separability	es
dc.subject	Landsat-8	es
dc.subject	Random tree ensembles	es
dc.subject	Random forest	es
dc.subject.other	CDU::9 - Geografía e historia	es
dc.title	Isolation forests to evaluate class separability and the representativeness of training and validation areas in land cover classification	es
dc.type	info:eu-repo/semantics/article	es
dspace.entity.type	Publication	es