Ir al contenido

Documat


Knowledge-Based Cataloguing of Deep Multi-Wavelength Cosmological Surveys

  • Autores: María José Márquez Sánchez
  • Directores de la Tesis: Luis Manuel Sarro (dir. tes.) Árbol académico, Tamás Budavári (dir. tes.) Árbol académico
  • Lectura: En la UNED. Universidad Nacional de Educación a Distancia ( España ) en 2016
  • Idioma: español
  • Tribunal Calificador de la Tesis: Francisco Javier Díez Vegas (presid.) Árbol académico, Miguel García Torres (secret.) Árbol académico, Emmanuel Bertin (voc.) Árbol académico
  • Enlaces
  • Resumen
    • I developed an expert data processing pipeline to improve the quality of the astronomical multi-colour galaxy catalogues. This pipeline acts in three main areas: (1) rening surface brightness measurements, (2) agging contaminated sources, and (3) improving the crossmatching of galaxies by modelling their colours. In order to develop this expert system, we start by building the knowledge model of entities and tasks involved and, as the result of this, we populate an ontology with instances of these classes. This ontology was created in line with the existing ones of the International Virtual Observatory Alliance(IVOA) initiative which intends to unify the Astronomy data models used across the scientic community. We achieved this goal through the use of articial intelligence methodologies. We used Bayesian model decision, active contour from Articial Vision and rule-based systems, all three of these methodologies are integrated into a single modular expert system versatile and exible enough to be used partially or entirely as a module of another processing system, as well as an input for a rened IVOA ontology. We performed a validation of the system with synthetic data and real astronomical images and catalogues from the Cosmic Evolution Survey (COSMOS). The outcome of this step conrmed the validity of our proposal. The results on a reduced but representative COSMOS data set conrmed that the photometric cross matching solution and the labelling of sources to determine their contamination status is a signicant improvement with respect to the existing methodologies. Regarding the renement of source contours, the experiment conducted with synthetic data as well as the results on real data demonstrated that this methodology is more exible in the determination of contours and it allows a good degree of automatism; however the computational cost and the observed lack of certainty for all cases indicates that this proposal needs further development. Overall it can be concluded that our system constitutes a significant advance in the cataloguing of galaxies in multi-band image cubes. We achieved this with an expert system that can be run without much human supervision.


Fundación Dialnet

Mi Documat

Opciones de tesis

Opciones de compartir

Opciones de entorno