Attitude Correlated Frames Based Calibration Method for Star Sensors

Liheng Ma, Shenglong Xiao, Wenjun Tang, Xiao Luo, Su Zhang2024 · DOI

TL;DR

This paper introduces a new calibration method for star sensors called the Attitude-Correlated Frame-Based Calibration Method (ACFCM). It addresses limitations in existing methods by combining the accuracy of laboratory calibration with the practicality of outdoor observation, leading to significantly improved principal point calibration repeatability and reliability.

What Problem Does This Paper Address?

Star sensors are crucial for precise attitude determination in spacecraft and other applications, but their accuracy relies heavily on precise calibration. Traditional calibration methods fall into two main categories: attitude-dependent (requiring expensive simulators and rotation tables) and attitude-independent (like the Interstar Angular Invariance Calibration Method - IAICM, which is simpler but suffers from unstable principal point calibration and insufficient star data). This paper tackles the challenge of achieving high-accuracy star sensor calibration without relying on costly equipment, while also improving the stability and reliability of key calibration parameters, particularly the principal point.

Methods

The proposed Attitude-Correlated Frame-Based Calibration Method (ACFCM) builds upon the strengths of existing approaches. It leverages outdoor star observations, similar to IAICM, thus eliminating the need for expensive laboratory equipment. The core innovation lies in correlating multiple star image frames captured at different times. This correlation is facilitated by a strapdown gyro unit, which provides attitude information. By combining these frames, the ACFCM effectively creates a larger, more uniform distribution of star images, which is crucial for accurate calibration. This approach increases the number of usable star images and improves the uniformity of their distribution, leading to more stable and reliable calibration results, especially for the principal point. The method then uses a non-linear optimization approach, specifically the Levenberg-Marquardt algorithm, to solve for the calibration parameters.

Datasets

The paper utilizes simulated data for initial testing and validation. For experimental validation, they conducted an outdoor star observation experiment in Dawei Mountain National Forest Park. The experiment involved a star sensor and a 50-type Laser Gyro Unit (LGU). They collected 10 groups of data, each containing 169 star image frames, with a total of 30,347 validly imaged stars used in the final analysis.

Key Insights and Ideas

  • Bridging the Gap: The ACFCM successfully bridges the gap between expensive, precise laboratory calibration and simpler, but less accurate, outdoor methods.
  • Gyro-Assisted Correlation: The use of a strapdown gyro unit to correlate frames is a key enabler, allowing for a richer dataset without complex mechanical setups.
  • Improved Principal Point Stability: The most significant insight is the dramatic improvement in the repeatability and reliability of the principal point calibration, a known weakness of IAICM.
  • Uniform Star Distribution: Correlating frames leads to a more uniform distribution of stars across the effective image, which is critical for accurate parameter estimation.
  • Reduced Reliance on Equipment: The method significantly reduces the dependence on expensive calibration equipment like precision rotation tables and star simulators.

Key Results

  • Simulation: In simulations with Gaussian white noise, the ACFCM achieved a principal point calibration error approximately one-third that of the IAICM method when the noise standard deviation increased to 0.2 pixels. The focal length and distortion calibration results were also superior.
  • Experiment: Experimental results showed that the ACFCM produced principal point calibration results with much smaller variations across different data groups compared to IAICM. The maximum difference for the principal point x0 was less than 0.8 pixels with ACFCM, versus 8.7 pixels with IAICM.
  • Residual Errors: The root mean square (RMS) of residual errors in the experiment was significantly lower for ACFCM (0.0057, 0.0055 pixels) compared to IAICM (0.0115, 0.0115 pixels), indicating higher calibration accuracy.

Limitations and Future Work

  • Gyro Error Consideration: While the paper states that gyro errors can be ignored for short correlation times (up to 500s for the 50-type LGU), the impact of gyro errors on longer correlation times or with less accurate gyros is not deeply explored.
  • Star Centroid Errors: The paper acknowledges that star centroid errors directly affect the accuracy of the observation vector and warrant further investigation.
  • Earth's Motion: While the impact of Earth's motion is deemed negligible for the specific star sensor's angular resolution and exposure time, the authors suggest that for scenarios with reduced star centroid errors, Earth's motion might need to be theoretically considered for enhanced accuracy.
  • Future Work: The authors mention that they have proposed a method for calibrating mounting angles in another paper, implying that this aspect could be further integrated or refined within the ACFCM framework. They also suggest further investigation into star centroid errors.

Summary updated

BibTeX
@misc{ma2024attitude,
  author = {Liheng Ma and Shenglong Xiao and Wenjun Tang and Xiao Luo and Su Zhang},
  title = {{Attitude Correlated Frames Based Calibration Method for Star Sensors}},
  year = {2024},
  doi = {10.3390/s24010067},
}

Loading related content...

Figure