On-Orbit Geometric Calibration from the Relative Motion of Stars for Geostationary Cameras

Linyi Jiang, Xiaoyan Li, Liyuan Li, Lin Yang, Lan Yang, Zhuoyue Hu, Fansheng Chen2021 · DOI

TL;DR

Geostationary remote sensing cameras (RSCs) often suffer from geometric inaccuracies after launch due to environmental factors. This paper introduces a novel on-orbit geometric calibration method that leverages the relative motion of stars to achieve high precision, overcoming limitations of traditional methods like cloud cover dependency and insufficient star availability. The proposed approach significantly improves positioning accuracy to within ±2.35 pixels.

What Problem Does This Paper Address?

Remote sensing cameras, especially those in geostationary orbits (GSCs), are crucial for continuous Earth observation. However, the harsh conditions of launch and the space environment can alter their geometric positioning models, leading to reduced image accuracy. Traditional on-orbit calibration methods rely on Ground Control Points (GCPs) or star observations. GCP-based methods are hampered by cloud cover and the difficulty of acquiring sufficient high-precision GCPs. Star-based methods, while less affected by weather, can be limited by the number of visible stars in a single image, making accurate calibration challenging. This paper addresses the need for a robust, weather-independent, and star-count-independent on-orbit geometric calibration method for geostationary cameras.

Methods

The core of this paper is a novel on-orbit geometric calibration method that utilizes the relative motion of stars. The approach involves:

  • Geometric Calibration Model Construction: A detailed geometric positioning model is established, considering the camera's optical system structure and its relationship with the satellite's coordinate system. This model incorporates interior (e.g., lens distortion, principal point) and exterior (e.g., attitude, orbit) parameters.
  • Relative Motion Analysis of Observed Stars: Instead of relying on a static view of stars, the method analyzes the trajectory of stars across multiple consecutive images. This is achieved by tracking the movement of stars as the satellite's pointing mirror adjusts.
  • Object Vector (OV) Generation: By observing the relative motion of stars, the method generates abundant Object Vectors (OVs). These OVs represent the direction of the star in the satellite's body coordinate system. The key insight here is that the apparent motion of a single star across multiple frames can be treated as if multiple different stars were observed simultaneously, thus overcoming the limitation of a small number of visible stars.
  • Iterative Calibration Parameter Estimation: The generated OVs are then used iteratively to correct the calibration parameters. This involves a stepwise approach: first, exterior parameters are estimated, followed by interior parameters within the reference frame established by the exterior parameters. This stepwise estimation helps reduce parameter correlation and improve accuracy.
  • Error Representation: The positioning error is quantified by comparing the celestial coordinates derived from the calibrated model with those from a star catalog. Errors are expressed in pixels for intuitive understanding.

Datasets

  • On-orbit observation data from a geostationary experiment satellite: The paper utilizes real star observation data from a geostationary satellite. Specific details about the satellite's parameters (orbit altitude, focal length, sensor information, pixel size, attitude measurement accuracy) are provided in Table 1.
  • Data Collection Period: Observations were collected over 20 consecutive days (August 2nd to August 21st) to account for potential daily variations caused by thermal deformation.
  • Test Data: For experiments, stellar trajectories with more than 25 track points were selected. Five track points per trajectory were randomly chosen for error testing, while the remaining were used for parameter calculation.

Key Insights and Ideas

  • Leveraging Stellar Motion for Abundant References: The most significant insight is the transformation of a single star's apparent motion into a source of multiple calibration references. This elegantly bypasses the problem of insufficient stars in a single frame, a major limitation of traditional star-based methods.
  • Weather Independence: By using stars, the method is inherently independent of cloud cover, a critical advantage over GCP-based calibration for Earth observation satellites.
  • Robustness to Thermal Deformation: The method implicitly accounts for thermal deformation by analyzing the changing relative positions of stars over time, which is a known source of error for geostationary satellites.
  • Stepwise Calibration for Improved Accuracy: The adoption of a stepwise calibration (exterior then interior) is a deliberate choice to reduce parameter correlation and enhance the accuracy of the estimated calibration parameters.

Key Results

  • Significant Improvement in Positioning Accuracy: The proposed method reduced initial positioning errors from approximately -6.795 pixels (RA) and -21.004 pixels (DE) to within ±0.85 pixels after calibration.
  • Final Accuracy within ±2.35 Pixels: The paper reports that the positioning errors, after correction, are concentrated and stably obey a normal distribution, with the probability of RAerror within ±2.24 pixels and DEerror within ±2.35 pixels at a 95% confidence level.
  • Influence of Track Points: Experiments showed that increasing the number of track points used for calibration generally improves accuracy, with the error stabilizing when more than 20 track points are utilized.
  • Effective Curve Fitting: The smoothing spline method used for fitting stellar trajectories effectively reduced errors caused by satellite jitter, pointing inaccuracies, and sampling errors, achieving fitting errors generally within ±0.03 pixels.

Limitations and Future Work

  • Random Measurement Errors: The authors acknowledge that random measurement errors, such as orbit and attitude measurement errors, can still influence the calibration accuracy.
  • Accuracy of Star Centroid Extraction and Trajectory Fitting: The accuracy of these preprocessing steps directly impacts the final positioning accuracy and is suggested as an area for future study.
  • Generalizability: While the method is proposed for geostationary cameras, the authors suggest it is likely suitable for other RSCs due to the similar spatial relative motion relationship between satellites and stars. Further validation on different satellite platforms would be beneficial.

Summary updated

BibTeX
@misc{jiang2021orbit,
  author = {Linyi Jiang and Xiaoyan Li and Liyuan Li and Lin Yang and Lan Yang and Zhuoyue Hu and Fansheng Chen},
  title = {{On-Orbit Geometric Calibration from the Relative Motion of Stars for Geostationary Cameras}},
  year = {2021},
  doi = {10.3390/s21196668},
}
Figure