Joint laboratory-field two-stage calibration method for airborne star sensor

A joint laboratory-field two-stage calibration method is proposed to address two critical problems in airborne star sensor calibration: (1) the high modeling order caused by minute systematic errors from the mirror-servo mechanism assembly and alignment errors, and (2) the estimation inaccuracy of the imaging model due to coupling between intrinsic and extrinsic parameters in laboratory calibration. In the first stage, under laboratory conditions, a 30-dimensional nonlinear model based on 2×Pitch physical features is designed using the law of reflection, and singular value decomposition (SVD) closed-form ridge regression is employed to establish the pixel-to-boresight mapping, thereby bypassing the high-dimensional coupling problem associated with explicit modeling of each mirror-servo mechanism assembly error. In the second stage, field stellar observations are utilized to refine the reference vector through Procrustes analysis and Nelder-Mead optimization, and a weighted joint refinement strategy fuses both data sources to circumvent the calibration inaccuracy caused by coupling between laboratory extrinsic and intrinsic parameters. Experimental results show that the proposed method achieves a laboratory root-mean-square residual of 1.48 ″ and reduces field residual to 12.47 ″~14.11 ″, representing approximately a 5.8× improvement over traditional geometric methods. Comparison with three baseline calibration methods demonstrates that the proposed method can effectively resolve the high-order modeling and parameter coupling problems in airborne star sensor calibration.