Abstract:
To address the critical challenge of on-orbit calibration for event cameras in demanding space applications, such as space situational awareness, on-orbit servicing, and visual navigation, this paper proposes a line-based calibration method that leverages the inherent structural features of spacecraft. Unlike conventional calibration approaches that necessitate specialized calibration patterns or artificial targets, our method eliminates the dependence on specific calibration boards or pre-deployed calibration infrastructure, thereby making it practical and feasible for real-world on-orbit deployment scenarios, where such auxiliary equipment is unavailable or impractical. The proposed framework leverages the unique characteristics of event cameras by first developing a robust line extraction algorithm specifically designed to handle the asynchronous and sparse nature of event streams. This algorithm can directly extract 2D lines from spacecraft event streams with high reliability, even under challenging lighting conditions and dynamic motion scenarios commonly encountered in space environments. Subsequently, given the known 3D geometric model of spacecraft, we establish rigorous geometric projection constraints between the extracted 2D observation lines and their corresponding 3D model lines. An initial estimation of both intrinsic and extrinsic camera parameters is then efficiently derived using the Direct Linear Transformation method, providing a robust foundation for subsequent optimization procedures. To further improve calibration precision, we formulate a nonlinear optimization objective function that minimizes the Euclidean distance between valid events and their reprojected lines. This approach enables joint refinement of camera parameters, including focal length, principal point, distortion coefficients, and extrinsic parameters, using nonlinear optimization algorithms. Comprehensive calibration experiments, encompassing both simulation studies and ground-based testing, demonstrate that the proposed method achieves higher accuracy in critical parameters compared to state-of-the-art approaches, thereby validating its effectiveness and robustness under complex dynamic conditions. This work provides an efficient, flexible, and environment-adaptive solution for on-orbit event camera calibration, delivering reliable calibration parameters for subsequent pose estimation, satellite tracking, and other vision-based space applications.