| In the study of high-precision astrometry,this research group needs to process the CCD astronomical image first.The astronomical image processing software developed by the research group will generate a data file after pre-processing,star search,centering and matching,which will be used for the subsequent reduction process.Specifically,we use the least squares fitting method to solve the plate constant model of the reference star in the data file,that is,to solve the polynomial relationship between the theoretical position(standard coordinates)of the star image and the measured coordinates of the pixel position on the CCD image.In general,the residuals of the position measurements of reference star are relatively small.However,sometimes there are stars with large measurement errors.If they are not processed,the solution of the whole model will be affected and the measurement results of all stars will be affected,resulting in the inaccuracy of the final measurement results.Therefore,in the process of data processing,we usually need to eliminate the influence of these abnormal stars to ensure the best measurement results.The reweighting scheme is a universal scheme for improving chi-square minimization.It was first proposed by Stetson,a Canadian scientist,and applied to stellar photometry in history.In this paper,we extend it to the high-precision position solution of the target star in the CCD astronomical image data,and compare it with the random sampling consistency algorithm on this basis.The reweighting scheme and random sampling consistency algorithm are applied to the processing of abnormal stars.In the research process of this paper,the precious data of a batch of CCD astronomical images measured by the research group were simulated and processed.For the simulated data of stars with large errors,the influence of the large position errors of these stars on the measurement results of target stars was reduced by using a parameter weighting scheme,rather than simply eliminated.The reweighting scheme was used to solve the plate constant model for simulation data under different conditions,and the mean O-C(observed value-calculated value)and precision(standard deviation)of the target star before and after the use of the reweighting scheme were observed.The improvement was evaluated by comparing the mean O-C and the precision with the measurement results from the original data.In addition,we also compare the improvement effect of the random sampling consistencyalgorithm under the same conditions.The results show that the reweighting scheme is an effective and reasonable scheme to deal with data sets containing a certain proportion of abnormal data points.For the simulation data with noise(Gaussian noise with mean value of0.1～0.5pixel and standard deviation of 0.15～0.25pixel)ratio of 10% to 30%,the reweighting scheme can improve the measurement results of the target star of the simulation data.For the simulation data with noise ratio of 20%,the improvement effect is the best,and it is very close to the measurement results of the original data.In addition,when the target star deviates from the center of the field of view,this scheme can also play an improving role.Our experiments show that the best improvement effect is achieved when the two parameters ? and ? in the reweighting scheme are 1.5 to 2.0 and 4.0 to 8.0 respectively.Finally,compared with the random sampling consistency algorithm,the reweighting scheme is a more effective and robust method. |
