In depth studying the phenomenon of solar atmospheric activity requires large field of view, long time and multi band high spatial and temporal resolution data. At present, almost all large aperture solar telescopes are based on adaptive optics technique and high resolution image reconstruction technique to realize the high spatial resolution observation of the sun. High resolution image reconstruction can effectively overcome the influence of atmospheric turbulence and make solar telescope achieve near diffraction limit resolution imaging. Compared with the adaptive optics technique, it is easy to operate and realize, but also lack of real time that because of the computing is time-consuming and reconstruction process are cumbersome and complex.With one meter New Vacuum Solar Telescope NVST was successfully completed and put into use in 2010, the domestic solar observations goes into the era of high spatial resolution. NVST is currently using speckle reconstruction method to deal with daily observation data. In order to deal with massive data in a timely and rapid manner, we must have an efficient data processing strategy and high speed reconstruction algorithm. In this thesis, based on the NVST actual observation, we study on the high speed high resolution solar images reconstruction algorithm. The research contents include the NVST data reconstruction model, the optimization of the reconstruction algorithm and the parallel operation of the algorithm. Details of the study and the corresponding progress achieved are as follows：According to the actual observation and data processing status, we establish data processing mode Level1 and Level1+. The two level processing mode allows NVST to release data in a timely manner in the case of speckle reconstruction algorithm is time consuming. Under the precondition of ensuring the accuracy of data processing, we analyze and optimize the shift-and-add method and the speckle masking method. We analysis the key reconstruction parameters through experiments, and simplify the computation as much as possible to improve the efficiency of data reconstruction.On the basis of the optimization of the algorithm, we use hardware to improve the speed of the high resolution reconstruction. We discuss the part of the high resolution reconstruction algorithm that can be executed in parallel computing, and perform experiments of high resolution reconstruction algorithm accelerating in the GPU and cluster. We realize the real-time reconstruction of class Level1 in GPU, and realize the parallel computation of sub block image four-dimensional bispectrum. In addition, we also achieve rapid reconstruction of Level1 and Level1+ in cluster.Application of Level1 and Level1+ to the practical processing of NVST multi-channel data is studied, and solution to the problems encountered in the actual observation and data handling is given. We propose a method to eliminate the influence of the fixed image noise generated by the sCMOS. According to the results of the Level1+ reconstruction of the moon, it is seen that the “second flat-fielding” can effectively suppress the fixed noise amplification in the speckle masking reconstruction. An improved cross correlation method is proposed for the reconstruction of the solar limb. This method can effectively avoid the correlation error, and guarantee the smooth progress of the reconstruction of solar limb activity. To improve the signal to noise ratio of the chromosphere data, we adopt and improve the Level1 reconstruction algorithm. The resolution of the reconstructed result is greatly improved.
修改评论