On-Orbit Auto-Focusing Method of Space Camera Based on Multistar Image

To enhance the efficiency and accuracy of the on-orbit focusing of Earth observation satellites, this study proposes a method based on multistar imaging. Specifically, the proposed method quantitatively evaluates the defocusing state of a space optical camera and estimates the optimal position of its photosensitive surface of detector (PSD). An experimental platform is established to capture star images, and the performance of the algorithm is evaluated based on indexes such as focusing accuracy and robustness. Moreover, the algorithm is compared to the single-star-imaging-based algorithm. Consequently, the effectiveness of the proposed algorithm is validated. Next, to demonstrate the application of the algorithm to real-time on-orbit focusing, field programmable gate arrays (FPGAs) are used. Specifically, the performance of the multistar-imaging-based algorithm in practical engineering applications is verified using space camera on-orbit star imaging data, and its performance is compared with that of the single-star-imaging-based algorithm. In ground tests, the maximum focusing error of the algorithm based on multistar imaging is 0.00075 mm. During robustness tests involving variations in the focusing step, star centroid distribution uniformity, and image noise, its maximum error is 0.0021 mm. Further, when the signal-to-noise ratio (SNR) of the image exceeds 20 dB, the focusing accuracy and goodness of fit (GOF) of the algorithm based on multistar imaging are better compared to those of the single-star-imaging-based algorithm. The prototype verification results present an error of 1.623% relative to the baseline data. Using on-orbit star imaging data, the error between the optimal PSD position and the actual ideal PSD position is calculated to be 0.01528 mm. These results indicate that the auto-focusing algorithm proposed in this article demonstrates high focusing accuracy and robustness against external factors such as focusing step changes, centroid distribution uniformity variations, image noise differences, and star brightness changes. Moreover, it outperforms the single-star-imaging-based algorithm. Hardware deployment verification reveals that the computing resources and calculation accuracy of the algorithm within the XC7K325T environment satisfy operational requirements. Thus, the proposed algorithm can be employed for the on-orbit focusing tasks of space cameras, satisfying image quality requirements with enhanced focusing accuracies.