Técnicas automatizadas de detección y análisis del efecto de lente gravitacional en cartografiados de galaxias de nueva generación

Abstract

This thesis has focused on the study of the gravitational lensing effect, and the development of automated techniques for identifying this effect in astronomical images produced by future surveys, such as J-PAS and Euclid. First, a multi-frequency detection technique (matched filter) has been adapted and implemented on astronomical images and already-identified sources from the Herschel-ATLAS survey. Among these objects, we have found both very weak sources, for which the method has improved their signal significance, and possible lensed systems candidates. Secondly, we have estimated, through an analysis of the gravitational lensing effect, the distribution of dark matter in a cluster of galaxies as well as the position of the multiply-lensed sources. Finally, several models of convolutional neural networks have been trained for the detection of the gravitational lensing effect in simulated images of KiDS, Euclid, and J-PAS.