Research And Realization Of Real - Time Acquisition And Distributed Storage System Of Sun Magnetic Field Image

This thesis is based on requirements of Huairou Solar Observing Station, National Astronomical Observatories, Chinese Academy of Sciences. It includes two part:First, the design and implementation of a new real-time correlation solar observing system based on GPU for acquiring the deep-integration magnetogram; Two, the design and implementation of a massive solar image data storage. Both systems have been put into use at HSOS.Sun’s magnetic field observations generally require multi-frame monochromatic images superimposed to obtain a higher magnetic field sensitivity of the image, and this way you need to collect observations to each frame displacement related operations. The method is usually used in hardware such as an adaptive optics and correlation tracker, and some of the history of the telescope and not enough room for such changes, it is necessary to use correlation calculation software manner. But with the CCD (charge-coupled device) hardware efficiency of existing observing systems have been unable to keep up with the CCD acquisition speed. Text design and implement a new generation of solar magnetic field system based on real-time observation GPU (Graphics processing unit) for parallel computing, the system redesign and expand the functionality of the original system, to achieve a real-time image acquisition overlay displacement parallel processing. Mainly related to work is as follows:First, through CUDA (Compute Unified Device Architecture) framework for the use of GPU parallel computing to achieve a displacement stack algorithm, greatly improving the efficiency of the algorithm, to meet the real-time requirements of the relevant observations. Experimental analysis of each step of the algorithm superimposed displacement and targeted optimization.Second, the expansion and redesign of the entire observation system provides an interactive overlay area selection displacement and other related functions, increasing the ease of use of the system.On the other hand, the traditional centralized data storage has been unable to meet the modern solar observations massive data storage requirements. This article describes the National Astronomical Observatory of Huairou Solar Observing base (HSOS) to achieve, based on the Hadoop distributed data storage system. It implements the massive solar observation data safe and secure storage, management, retrieval, reading and maintenance, and the dynamic growth and to meet the data requirements of distributed batch processing of stored data. The main contents are as follows:First, according to the characteristics of the design of the solar image data storage platform for large-scale solar image, introduced a new data storage structure. Use reliable distributed storage technology to store massive solar observation data, read, management and expansion, to ensure the consistency and security of data.Second, to achieve the various features of the platform, including the preservation of a dozen different forms of intelligent analysis of historical data upload, Web form data quickly display and query, batch download images and simple sun image distributed computing framework.In this stage, the two systems have been put into use in Huairou Solar Observing Station. Real-time observations of solar magnetic field system has been running stable for over a year, get a lot of high quality magnetic image, and the work has been published in an SCI journal articles as the first author. Large-scale solar image storage system has completed archive Ha telescope nearly30years observation data, and provides query and download services of those data.