Date of Award

5-2023

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Aerospace, Physics, and Space Sciences

First Advisor

Csaba Palotai

Second Advisor

Xianqi Li

Third Advisor

Hamid Rassoul

Fourth Advisor

David Fleming

Abstract

Bolide recording and analysis are crucial so meteoroid fragments can be found, a lightcurve analyzed, and its trajectory calculated. The Spalding Allsky Camera Network (SACN) generates videos and composite images of the night sky that are potential meteors based on changes in brightness. The best way to ensure quick identification is to automate the detection of bolides (and all meteors) using computational techniques. This project tested three algorithms to sort events between those with and without meteors - a Traditional Hough Detection Method, Convolutional Neural Network (CNN), and YOLOv5 against the previous technique from 2018 by Elena Botella. All the methods improved from the Optimized Botella method, but none performed well enough to automate detection fully. The best-performing method on all the metrics is the naive approach to the CNN, but it misses big, bright meteors and bolides. The three new methods can be combined to leverage their strengths to automate meteor detection fully.

Comments

Copyright held by author

Download

Share

COinS