Space Object Detection and Monitoring Using Persistent Wide Field of View Camera Arrays

Date of Award


Degree Name

M.S. in Computer Engineering


Department of Electrical and Computer Engineering


Vijayan Asari


Automated monitoring of low resolution, deep-space objects in wide field of view (WFOV) imaging systems is an important and emerging technology for Space Domain Awareness (SDA). SDA involves the holistic process of monitoring and characterizing space objects in order to ensure a safe environment for satellite operations and employment. With the proliferation of satellites, referred to as 'Resident Space Objects' (RSOs), in all orbits, SDA requires WFOV optical sensors to detect and track the growing population of multiple low-light objects. The PANDORA sensor array, located in Maui at the Air Force Maui Optical and Supercomputing Site, is an exemplar of a scalable imaging architecture that can detect dim deep- space objects while maintaining a WFOV. The PANDORA system captures 20?×120? images of the night sky oriented along the GEO belt at a rate of two frames per minute. The PANDORA sensor system makes possible the passive monitoring of hundreds of RSOs, but requires advanced image processing and exploitation techniques to autonomously and reliably be utilized. This thesis explores image processing and deep learning techniques to exploit PANDORA sensor data for use in SDA. To benchmark object detection performance, a synthetic dataset and annotated physical dataset of PANDORA imagery is prepared. Classical feature- based object detections are explored, which are tailored to specific space object morphologies in PANDORA imagery. Single frame object detection performance with developed classical methods are evaluated on the synthetic PANDORA dataset. Deep learning object detection techniques are then employed, which set a standard for WFOV low-resolution object detection. We present a deep learning RSO detection and tracking architecture: PASTOR (Persistent All- Sky Tracking and Object Re-Identification). This architecture consists of a deep-learned object detector using YOLOv5, with an object tracker consisting of Kalman filters. We present detailed analysis of PASTOR object detection and tracking performance on the physical annotated PANDORA dataset, reporting a maximum F1 score of 0.814, corresponding to 0.766 precision and 0.868 recall. With a high accuracy in object detection and tracking, PASTOR can be deployed to the PANDORA sensor to autonomously and reliably monitor hundreds of space objects simultaneously, whose exploited data can be leveraged for satellite orbit catalog maintenance and space-object anomaly detection.


Computer Engineering, Computer Science, WFOV, Pandora, SDA, RSO, satellite detection, satellite monitoring, deep learning, space force, space domain awareness, all-sky, EO astronomy

Rights Statement

Copyright © 2022, author