Construction of the Guide Star Catalog for Double Fine Guidance Sensors Based on SSBK Clustering

Yuanyu Yang, Dayi Yin, Quan Zhang, Zhiming Li2022 · DOI

TL;DR

This paper introduces a novel method, Spherical Spiral Binary K-means Clustering (SSBK), to construct a highly uniform and efficient guide star catalog for space telescopes. By combining global spherical division with local binary clustering, SSBK significantly reduces catalog size while ensuring a high probability of finding suitable guide stars, crucial for precise attitude determination.

What Problem Does This Paper Address?

Precise attitude determination is fundamental for space telescopes to achieve their scientific objectives. Fine Guidance Sensors (FGS) are critical components that enable this precision by tracking stars. The performance of FGS heavily relies on the quality of the guide star catalog used for star identification. Existing catalogs often suffer from non-uniform star distribution, leading to either an excessive number of stars in dense regions (increasing processing load) or insufficient stars in sparse regions (risking identification failure). This paper addresses the need for a guide star catalog that is both uniformly distributed across the celestial sphere and compact in size, thereby improving star identification speed and attitude determination accuracy for missions like the Chinese Survey Space Telescope (CSST).

Methods

The core of this paper is the proposed Spherical Spiral Binary K-means Clustering (SSBK) algorithm. It comprises two main stages:

  1. Global Uniform Division (Spherical Spiral Reference Point Method): This method, adapted from Bauer et al., divides the entire celestial sphere into regions of uniform spacing. It uses a spiral pattern to ensure that the distance between adjacent spiral circles and points along the spiral is consistent, creating a globally uniform grid of reference points. These points serve as centers for subsequent local analysis.
  2. Local Clustering (Binary K-means Clustering): For each region defined by the spherical spiral reference points, a binary K-means clustering algorithm is applied to the stars within that region. This algorithm iteratively divides clusters into two, minimizing the sum of squared errors (SSE) at each step. The process continues until a desired number of clusters (N) is reached, or until clusters become too small to subdivide. The brightest star within each final cluster is then selected as a guide star. This approach aims to create locally dense clusters of guide stars where needed, while still maintaining overall uniformity.

The paper also details essential preprocessing steps for the source catalog:

  • Star Pre-screening: Removing non-stellar objects and variable stars that could interfere with identification.
  • Magnitude Correction: Adjusting magnitudes to account for differences in sensor spectral response.
  • Position Correction: Transforming star positions from catalog epoch to observation time, accounting for proper motion, parallax, aberration, and light deflection.
  • Double Stars Elimination: Identifying and removing close binary stars that could be mistaken for a single star, impacting accuracy.

Datasets

The primary dataset used in this research is the Gaia Data Release 2 (GaiaDR2). This catalog contains approximately 1.7 billion stars with precise positional and photometric data. The paper utilizes GaiaDR2 as the source catalog due to its high accuracy and comprehensive coverage, which are essential for constructing a high-precision guide star catalog. No new benchmark datasets were introduced; instead, GaiaDR2 was preprocessed and filtered to create the candidate and final guide star catalogs.

Key Insights and Ideas

  • Bridging Global and Local: The SSBK algorithm's strength lies in its dual approach: ensuring global coverage and uniformity through the spherical spiral method, and then optimizing local star density and distribution using binary K-means clustering. This addresses the limitations of methods that focus solely on one aspect.
  • Uniformity is Key for Efficiency: The paper strongly emphasizes that a uniform distribution of guide stars is crucial for both efficient star identification (avoiding dense regions) and reliable attitude determination (ensuring enough stars are always visible).
  • Compactness Matters: By reducing the total number of stars in the catalog significantly (over 91% reduction from the candidate catalog), SSBK makes star identification faster and reduces storage requirements, which is vital for onboard processing in space missions.
  • The "Sweet Spot" for Star Count: The research highlights that for high-precision FGS, a number of stars in the Field of View (FOV) between 5 and 15 provides the best balance for high identification probability and attitude accuracy. SSBK excels at achieving this range.

Key Results

The SSBK algorithm was evaluated using a simulated FOV of 0.2° × 0.2° and a magnitude range of 9-15 mag, with a threshold of 9 stars per FOV.

  • Catalog Size Reduction: The final guide star catalog contains 9,298,862 stars, a reduction of 91.09% compared to the initial candidate star catalog (104,350,738 stars).
  • Uniformity: The SSBK-based catalog demonstrates significantly better uniformity than the candidate catalog and other compared methods (MFM, SSRP). This is evidenced by a lower standard deviation of the number of stars per FOV.
  • Probability of Optimal Star Count: The probability of having 5-15 stars in the FOV is exceptionally high: 99.39% for single FOVs, 99.41% for FGS1, and 99.59% for FGS2. This is substantially higher than other methods.
  • Reduced Average NOS: The average number of stars (NOS) in the FOV is maintained at approximately 9, which is considered optimal for high-precision FGS.

Limitations and Future Work

  • Generation Time: The paper acknowledges that the large number of spherical spiral reference points and the iterative nature of the binary K-means clustering algorithm lead to a long generation time for the guide star catalog.
  • Real-time Application: While the constructed catalog is promising for high-precision attitude determination, future work should focus on integrating it with real-time star identification and attitude determination algorithms to enable onboard calculations. This will likely involve optimizing the catalog generation process or developing faster algorithms for real-time use.

Summary updated

BibTeX
@misc{yang2022construction,
  author = {Yuanyu Yang and Dayi Yin and Quan Zhang and Zhiming Li},
  title = {{Construction of the Guide Star Catalog for Double Fine Guidance Sensors Based on SSBK Clustering}},
  year = {2022},
  doi = {10.3390/s22134996},
}
Figure