Feasibility of Event-Based Sensors to Detect and Track Unresolved, Fast-Moving, and Short-Lived Objects
Date of Award
M.S. in Electrical Engineering
Department of Electrical and Computer Engineering
Event-based sensors are an emerging technology that hold the potential to change the computer vision field. Specifically, event-based sensors have increased dynamic range, increased temporal resolution, and lower power consumption. These qualities give a distinct advantage over traditional framing sensors in a variety of computer vision applications including high speed imaging, feature extraction and detection, etc. Event-based sensors have just recently begun to be tested in the context of space situational awareness, namely with the detection and tracking of slow moving objects (i.e., stars and satellites). The feasibility of event-based sensors to detect and track fast-moving, short-lived, and unresolved objects in the context of space situational awareness is a fairly novel approach with limited supporting literature. This question is investigated through the context of meteor detection and tracking with the purpose of pushing the event-based sensor capabilities to their limit. Two data collections are performed and a data set containing both meteors and distractors (i.e., stars, airplanes, etc.) is captured. A processing pipeline is developed to allow for denoising of raw event data and subsequent detection and tracking of meteors. A simple tracker with several thresholding methods is chosen to detect and track the meteors and is found to be the most computationally inexpensive method while providing reasonable results. A calibration process is also proposed to explain the logarithmic relationship between intensity and the generation of events. Lastly, the detection and tracking method along with calibration is compared to the hand-annotated ground truth obtained in the data collection to determine accuracy. Overall the simple tracker results in a majority of meteors being detected and accurately tracked over their entire duration. The calibration method also provides an accurate transformation from event magnitude to traditional pixel intensity.
Astronomy, Electrical Engineering, Engineering, Technology, event-based sensors, event cameras, space situational awareness, detection and and tracking, sensor hardware, intensity profile, meteor detection and tracking, meteor data set, novel data set, novel sensor
Copyright © 2022, author
Tinch, Jonathan Luc, "Feasibility of Event-Based Sensors to Detect and Track Unresolved, Fast-Moving, and Short-Lived Objects" (2022). Graduate Theses and Dissertations. 7089.