Preprocessing of Astronomical Images from the NEOWISE Survey for Near-Earth Asteroid Detection

Date of Award

Summer 5-5-2021

Degree Type

Thesis

Department

Electrical Engineering

First Advisor

Dr. José Manjarrés

Second Advisor

Dr. Stephen Case

Abstract

Asteroid detection is a common field in astronomy for planetary defense which requires observations from survey telescopes to detect and classify different objects. The amount of data collected each night is increasing as better designed telescopes are created each year. This amount is quickly becoming unmanageable and many researchers are looking for ways to better process this data. The dominant solution is to implement computer algorithms to automatically detect these sources and to use Machine Learning in order to create a more efficient and accurate classifier. In the past there has been a focus on larger asteroids that create streaks in the telescope images, but this leads to less accuracy on classifying smaller and slower moving asteroids, which do not appear as often as streaks. We use image and source data from the NEOWISE survey telescope to create a preprocessing pipeline that allows for better training and testing data for a classifying Machine Learning algorithm. We were able to create a pipeline that finds sources on an image within .02 degrees as well as a collection of data on the known sources that can be used to set better parameters for finding specific sources in new images. These steps should aid in creating an algorithm focused on detecting smaller and slower moving asteroids, but our research also indicates that these may be more often mistaken for comets.

Comments

This poster describes work completed for Pence-Boyce summer research.

This document is currently not available here.

Share

COinS