Improved astrometric calibration of Cassini narrow angle camera images and its accurate geometric distortion model

The astrometric calibration of images taken from the China Space Station Telescope (CSST) is the key technique for preparing the CSST, which should start scientific operations around 2024. Taking the astrometric calibration of Cassini ISS images as an example, we hope to provide a reference for the images taken from the CSST. Observations of Cassini narrow angle camera with different orientations and pointings at dense star fields over five epochs (epochs 2006, 2008, 2010, 2014, and 2016) are used to perform astrometric reduction and to derive accurate geometric distortion models. For each frame, the star image centroids from the modified moment method are matched with the newest Gaia DR3 catalog and calculated to their topocentric apparent positions. Then, we reduce the observations using a fitted weighted fourth-order polynomial plate model and correct the effect of the charge transfer efficiency (CTE) issues (if they exist). To derive the geometric distortion model at each epoch, a weighted four-parameter plate model is used to calculate and derive the (O-C)s (the observed minus computed) of stars. We calculate the median of the (O-C)s in each area of a divided 8 x 8 box in the pixel coordinate after considering the effect of CTE. Finally, we fit the geometric distortion model at each epoch. Compared with the reduction results of Zhang et al. (2021), our results show a considerably improved precision of the stars. For brighter stars (e.g., 11-12 mag with an exposure time of 2.6 s), the standard deviations of the mean (O-C) are similar to 0.04 and similar to 0.03 pixels (approximately a 0.5- and 1.0-fold improvement, respectively) on the x- and y-axes, respectively. The geometric distortion models are stable over the five epochs. In the 1024 x 1024 pixel image, the maximum, median, and minimum of the geometric distortion values are 0.184 +/- 0.023, 0.038 +/- 0.005, and 0.007 +/- 0.003 pixels, respectively, which differ from Owen's model in 2003. The mean geometric distortion model, as well as the correction method, is given in this work.