Detecting variable astronomical objects over time images. Detección de objetos variables en imágenes astronómicas a través del tiempo.

Ana Bertha Cruz Martínez has finished her M. Sc. Thesis (CIC, 2016. In Spanish). It can be dowloaded from here.
Abstract
     One interesting and unusual phenomenon that the astronomy studies,  is the explosion of a supernovae. It has a short duration, on the order of days. The study of supernovae has become important because they are objects that have a known light curve, specifically the Ia supernovae type. These can be used as benchmarks to measure the acceleration of the universe’s expansion. Measuring the redshift of a supernova we could known how fast away the universe expands.
     To accomplish this task we have developed different types of tools and use techniques for automated observation data acquisition that have produced large volumes of information.
     Astronomers have begun the process of extracting information with computational techniques as an alternative to the systematic analysis of the data. The main techniques used include the image and signal analysis.
      This work proposes a model that can process images, implementing some signal analysis techniques that are useful for the treatment of time series, which astronomers call light curves. Supervised classification module was also implemented for three classes: stars, variable objects and supernovae candidates, and additionally a sub-classification of the supernovae candidates in their different types of supernova: I y II.
     The model was built experimentally, trying different methods for each stage of the process, some of these methods are handled in astronomy separately to process data and make detection of supernovaes and other phenomena, the developed modules are:
     • Preprocessing: photometry and labeling images with PPP software (Picture Processing Package) developed by Phd. H. Yee carrying out this task was integrated.
     • The creation of the database was automated generating and using different SQL Statements and MySQL as manager of the database, organizing by regions and catalogs.
     • Construction of time series of each object in the database like a tuple.
     • Treatment of time series: clean background noise, interpolation supernovas filter to soften the light curve and standardization.
     • Supervised classification: constant objects, variable objects and candidates. Subclassification in supernova types using real examples from the literature.
     The software was developed with Python, testing each module with two regions of the database project CHASE Observatory from Cerro Tololo, Chile. Subsequently, the entire database is processed and its distribution of the objects was determined in three classes. This work presents the proposed model, the developed software, the final results of classification and the set of supernovae candidates, and their classification in their different types.