Michael Stewart, York University; Regina Lee, York University; Shane Ryall, Defence Research Development Canada
Keywords: Image Processing, Python, Resident Space Object
Abstract:
Space Situational Awareness (SSA) is defined as the practice of tracking and measuring objects for the concerns of defence. In order to establish SSA, measurements of space fairing objects termed Resident Space Objects (RSOs) are collected with various sensors. Optical sensors offer flexibility for SSA measurements due to the matured nature of light measuring techniques matched with the ability to resolve the location of the RSO in relation with the background stars. Like all data, optical data requires processing to convert the raw data to usable scientific measurements. As most image processors contain proprietary code, the question is posed: “What are the steps required to develop a general-purpose image processor for SSA?”. The simplified answer is to use published methods for optical astronomic data reduction, photometric measuring methods, apparent magnitude transformations, astrometric techniques, and orbit determination algorithms to create an image processing pipeline. At DRDC this pipeline was developed and implemented in Python due to the language’s ease of use and availability of open-source packages. The purpose of this paper is to highlight each step in the image processing pipeline and to conclude with an updated orbit of targeted RSO’s as well as a calibrated photometric light curve from gathered optical images to contribute to SSA.
A simulation on the observational viability of the RSOs were conducted before each observation run. These simulations took in the location of the sensors, their field of view and look angles. From this, the simulator predicted the number of catalogued stars above a given magnitude that would be present in the frame. The requirements for astrometric measurements (determining the topocentric right ascension and declination) and subsequently orbit determination rely on four separate stars in the background of each image. Likewise, differential photometry; a form of photometry relying on the catalogued values of star in the background of image of the RSO requires several stars to compare to the RSO in the same image. Depending on the wishes of the user, absolute photometry may be performed if differential photometry is not possible. Absolute photometry requires observations of Landolt star fields with a large range of air masses throughout a given night to calculate the transformation required to convert instrumental magnitudes into apparent magnitudes. The specific Landolt star fields of SA 23 and SA 32 for absolute photometry was chosen based on its declination, latitude of the observation, and local sidereal time producing a range of airmasses for a given night.
After the schedule had been created, observations with a computerized mount and scope were used to capture the target RSOs that were considered viable. The target RSO was tracked using its latest Two-Line Element set (TLE) propagated to the observation time, obtained from celestrak.org. Along with these observations, bias, dark and flat frames were acquired. A suite of software was used to control this system consisting of TheSkyX, SkyTrack, CPWI and ASCOM. Observation campaigns in March 2022, August 2023 and January 2023 were conducted at the DRDC Ottawa Research Centre to capture RSOs of interest. The campaigns included the use of a Celestron 8″ telescope paired with a ZWO ASI 6200MM Pro (Mono) with the ZWO LRGB filter set. Objects of interest included IntelSat 10-02 with MEV-2 and Anik F1. These collected datasets were used to test the developed pipeline.
The image processing would start with image calibration by reducing the raw images using combined master bias, dark and flat frames. The sky background was then estimated and subtracted from the reduced images allowing the light sources in the image to be detected better. A light source detection algorithm was implemented, and the source’s centroid locations were sent to nova.astrometry.net service to plate solve the image. Once the World Coordinate System (WCS) information was resolved through plate solving, it was used to query the APASS star catalogue in the area of sky that the image looked at. Catalogue stars were matched to the image stars and transformations were calculated for use in either differential or absolute photometry depending on method chosen. The result of this process produced astrometric data (topocentric right ascension and declination) and the calibrated apparent magnitude light curve. Produced calibrated light curves of the sampled RSOs (Anik F1 and IntelSat 10-02) had apparent magnitudes ranging from 10 magnitude to 12 magnitude, within the expected magnitude range.
From the topocentric right ascension and declination of the RSO, the azimuth and elevation were calculated given the location of the observer (referred to as look angles). Since the sample RSOs were tracked with TLEs, prior knowledge of the orbits was known, so a Batch Least Squares algorithm with a given set of weights was used to update the orbit of the spacecraft. Preliminary results from astrometric analysis of the images shows that the angular difference between the measured and estimated (from the TLE) topocentric right ascension and declination is around 1.2 arcseconds, appropriate for preliminary results.
This paper further describes the application of each image processing step along with a brief theory. The updated orbit of the target RSOs along with its calibrated light curves is produced and analyzed further. In addition to this multiple photometry methods are compared to show their applicability to RSOs and Landolt star fields. Although this image processing is specific to continuous orbit determination with ground based sensors, an alternative deviation of the pipeline could be created for space based sensors and integrated on-board future mission for rapid orbital determination.
Date of Conference: September 19-22, 2023
Track: Space Domain Awareness