Aplication of elastic net algorithm in satellite images

• Autores: Príncipe, Erick, Willems, Bram
• Formato: artículo
• Fecha de Publicación: 2018
• Institución: Universidad Nacional Mayor de San Marcos
• Repositorio: Revistas - Universidad Nacional Mayor de San Marcos
• Lenguaje: español
• OAI Identifier: oai:ojs.csi.unmsm:article/20179
• Enlace del recurso: https://revistasinvestigacion.unmsm.edu.pe/index.php/fisica/article/view/20179
• Nivel de acceso: acceso abierto
• Materia: ENA; physical index; maximum likelihood; cluster; índica físico; máxima verosimilitud

This investigation was carried out in the Santuario Nacional Los Manglares de Tumbes (SNLMT), located in the district of Zarumilla department of Tumbes, is oriented to implement a methodology that allows to characterize the mangrove cover. For this, the image of the TM sensor of the LandSat 5 satellites was analyzed and processed evaluating a series of parameters related to the soil surface, such as SAVI (Index of vegetation adjusted to the ground), NDVI (Index of vegetation of dierence normalized) and NDWI (Normalized dierence water index) with a view to establishing the optimum index that allows to discriminate the dierent components of the Sanctuary's soil cover. The optimal index (SAVI) described above was introduced in the Elastic Net Algorithm (ENA) for the classication of the SNLMT ground cover. The images constructed from the ENA results were subjected to the validation process using conventional methods such as the maximum likelihood algorithm (MLA). This validation process consisted of performing the analyzes and comparisons of the average spectral signature graphs of each informational class obtained with both ENA and MLA, resulting in similar graphs where the RMSE was below 0.052 (dimensionless) and the Correlation factor on r=0.886. This indicates that the ENA method proves to be an eective tool for the subdivision of mangrove coverage classes.