A Comparison Of Centering Algorithms Based On CCD Images

With the development of the new observational methods and technique,it is more and more important to improve CCD image processing methods and raise its accuracy.Several classic numerical algorithms have been developed for centering celestial bodies.For instance,the one-dimensional centering algorithms include Gaussian fitting,Modified Moment method and median routines.While two-dimensional ones include Gaussian fitting,Modified Moment method.Of all these methods,the two-dimensional methods are used by the majority of astrometry researchers for its high-precision of measurement.For the CCD images from the actual observation,there is no unified conclusion of which is more precise among all these methods.In consequence,we need to improve the algorithms and achieve greater measurement accuracy in order to take the full advantage of CCD,while at the same time there is a necessity to compare the accuracy of all those methods in the same standards.Stone,van Altena and Auer studied and compared those Star-Centering algorithms based on photographic plates images or synthetically produced star images. Stone concluded that Modified Moment method was better than the others after he evaluated each of those centering methods in terms of processing speed,convergence properties,and the ability either to detect faint images or to deal with bright sky backgrounds,especially the measurement accuracy.A similar conclusion was also drawn by Ji K F through adopting synthetically produced star images and real CCD images.In this thesis,various classic centering routines were first realized either by their manual measurement or semi-automatic measurement.And then all kinds of algorithms were used to process the CCD images from the actual observations.At last the comparison of those methods based on the statistical results of different algorithm accuracies was made in order to find out a centering measurement algorithm with higher precision.CCD images adopted in this dissertation were observed by horizontal-1-meter reflecting telescope in the Xinglong Observational site,National Astronomical Observatories.In the processing of statistical analysis,we found that the imperfect of the telescope's mechanical system,which led to a distortion in the observation CCD field of view.Thus,a certain system error was caused by the imperfect of the telescope.An affine transformation was introduced to solve this problem.The data which reflects the accuracy of various methods is got after the elimination of systematic errors,after that these data is compared by the same procedures.After the comparison and analysis we can derive the measurement accuracies of various centering methods.The conclusion is that the one-dimensional centering algorithms are inferior to the two-dimensional centering algorithms,while Gaussian fitting is more accurate than two-dimensional Modified Moment method,and the advantage of Gaussian fitting is obvious in the process of the dark stars.