A comparison of centring algorithms in the astrometry of Cassini imaging science subsystem images and Anthe’s astrometric reduction

International audienceABSTRACT In the caviar software package, a standard tool for astrometry of images from the Cassini imaging science subsystem (ISS), Gaussian fitting is used to measure the centre of point-like objects, achieving a typical precision of about 0.2 pixels. In this work, we consider how alternative methods may improve on this. We compare three traditional centroiding methods: two-dimensional Gaussian fitting, median, and modified moment. Results using 56 selected images show that the centroiding precision of the modified moment method is significantly better than the other two methods, with standard deviations for all residuals in sample and line of 0.065 and 0.063 pixels, respectively, representing a factor of over 2 improvement compared to Gaussian fitting. Secondly, a comparison of observations using Cassini ISS images of Anthe is performed. Anthe results show a similar improvement. The modified moment method is then used to reduce all ISS images of Anthe during the period 2008–2017. The observed-minus-calculated residuals relative to the JPL SAT393 ephemeris are calculated. In terms of α × cos(δ) and δ in the Cassini-centred international celestial reference frame, mean values of all residuals are close to 0 km, and their standard deviations are less than 1 km for narrow angle camera images, and about 4 km for wide angle camera images