Star Elimination as a Means of Resident Space Object Identification for Space Situational Awareness

A fundamental tenet of Space Situational Awareness (SSA) is to properly identify resident space objects (RSOs) within unresolved optical space imagery. This subsequently allows the directional, i.e. right ascension (RA) and declination (Dec), data for each RSO to be processed resulting in an estimate of each RSO’s orbit. The emphasis here will be on developing an algorithm to identify Earth-orbiting RSOs that can be accomplished autonomously onboard a spacecraft. In order to distinguish RSOs from the star field, a suggested practice is to perform multi-frame image “collects” and identify RSOs based on how the scene changes from frame to frame. The RSO identification (RSOID) algorithm to be presented here is based on this philosophy and broadly consists of the following interdisciplinary steps: • Edge detection for each image in the collect, to determine the boundary between each set of illuminated pixels and the blackness of space; this reduces each image to numerous segmented “objects” (actually point spread functions) • Centroiding for each image in the collect, to localize each object in the camera frame based on (x,y) pixel coordinates • Registration (or “plate solving”) of each image, to determine the inertial pointing of the camera and thereby convert the (x,y) coordinates of each centroid in the camera frame to inertial coordinates, i.e. right ascension (RA) and declination (DEC) • Star elimination, to remove all centroids identified as belonging to a star, based on how closely clustered a group of centroids is throughout the different frames of the collect (i.e. standard deviation in RA and DEC) • Object association, to confirm the existence of each RSO “tracklet” (i.e. collection of RA/DEC data points) by performing linear regression on sets of centroids across the collect The algorithm requires the user to choose various thresholds at the start of a scenario. The primary threshold to be explored in this paper, termed the “star clustering” threshold, pertains to the distance between two centroids in different frames of a collect, in terms of RA and Dec coordinates. Centroids in different frames corresponding to the same star will be very closely spaced in RA and Dec, therefore a threshold can be defined such that if two centroids from different frames are closer than the threshold, they belong to the same star. This allows the two centroids to be eliminated from consideration. Ideally, if all stars in the collect can be eliminated, the “non-stars” that remain will consist of majority (if not all) RSOs. The RA and Dec points belonging to each RSO can then be joined together to form a tracklet. The RSO tracklets are, in fact, the desired output of the algorithm. The purpose of this research will be to develop and analyze a method that lends itself to autonomous operation onboard a spacecraft. Toward this end, the authors will create an adaptive means of choosing the “star clustering” threshold described above, the implication of which is to require less human-in-the-loop intervention. The algorithm is being tested on numerous real space image collects obtained via optical sensor configurations (i.e. telescope, mount, and CCD camera) available through the authors’ university. Initial progress includes generation of results with a baseline value for each of the above-described thresholds and exploring the effect of various threshold levels on the results. The results show that the ability of RSOID to correctly identify RSOs strongly depends on the value of each threshold, along with a variety of factors such as camera parameters and noise levels in the imagery. References to be cited in the paper will include: -T. Henderson, T. A. Lovell, D. Zuehlke, “Processing of Space Object Data from Optical Observers for Space Domain Awareness,” IEEE Aerospace Conference, 2021 -J. Du, X. Lei, “Space surveillance satellite for cataloging high-altitude small debris,” Acta Astronaut, Vol. 157, pp. 268-275, Apr. 2019. -B. Spanbauer, T. Kelecy, E. Lambert, “Automated near real-time validation and exploitation of optical sensor data for improved orbital safety,” Acta Astronautica, Vol. 157, pp. 404-414, Apr. 2019 -D. Zuehlke, “Space Image Processing and Orbit Estimation Using Small Aperture Optical Systems,” Embry-Riddle Aeronautical University, PhD Dissertation, Apr. 2019 -J. Machete, “Processing of Space Object Data from Optical Real-Time Ground-Based Space Debris Detection,” 1977 -T. A. Lovell, S. D. Zuehlke, T. Henderson, “An End-to-End Process for Local Space Situational Awareness from Optical Observers,” IEEE/ION Position, Location and Navigation Symposium (PLANS), pp. 1547-1555, 2020. -J. D. Anderson, T. A. Lovell, D. Zuehlke, A. J. Anderson, “Resident Space Object Identification in Arbitrary Unresolved Space Images,” AAS/AIAA Spaceflight Mechanics Meeting, 2023