Performance of star centroiding methods under near-real sensor artifact simulation

In this paper, robustness and accuracy of centroiding methods that are operating as part of star extraction process are investigated. Attitude determination systems consist of star extraction, star identification and pose estimation subsystems. However, most of the time the upper limit of their performance is determined by the accuracy of the star extraction step. In order to achieve higher localization accuracy under typical imaging conditions of near space, star extraction methods need to perform flawlessly during the localization of centroids in star images. In this work, a detailed setup for testing the robustness of star centroiding algorithms under typical working conditions is introduced. These conditions include inherent characteristics of the optical hardware and sensor electronics like spot of irradiance variation, readout noise and sampling errors, as well as, a source-based characteristic named photon noise. Performance evaluations of the compared methods are conducted based on the Centroid Estimation Error (CEE), which is a metric depending on the distance between the estimated and real star centroids.