【作者】 王翀；

【导师】 叶中付；

【作者基本信息】 中国科学技术大学 ， 信号与信息处理， 2010， 硕士

【摘要】 大天区面积多目标光纤光谱望远镜(Large Sky Area Multi-Object Fiber Spectroscopy Telescope)简称为LAMOST。它是我国在国家天文台兴隆观测站建立的一种大型光纤光谱天文望远镜。作为目前世界上光谱获取效率最高的光纤光谱望远镜,LAMOST既是我国天文学发展历史上的一个里程碑,也必将促进我国天文学在未来的长足发展。从本质而言,LAMOST也可以看成是一套复杂的成像系统,因此诸多的图像处理方面算法可以应用于其拍摄的二维光纤光谱图像处理。天文观测时所遇到的各种问题,都可以通过后期的图像处理来去除。本文所关注的就是在拍摄二维光纤光谱图像时所不可避免的宇宙射线干扰问题。本文调研了LAMOST的建设背景、观测模型以及数据处理流程。在后续的内容中,本文将重心放在对二维光纤光谱图像中一种特别的干扰——宇宙射线的处理上。首先对二维光纤光谱和宇宙射线的特性做了详细的分析,并提出了一种基于模板匹配的宇宙射线去除和修复算法。并在此基础上,提出利用基于GPU的通用计算编程方法来加速算法的处理速度,同时引入径向基函数(RBF)插值算法来改良图像的修复效果。本文主要做了以下几方面的工作:1.在概述二维光纤光谱图像特征的基础上,详细论述了宇宙射线的成因和在图像中的表现形式。根据它们在轮廓特征上的不同,提出了逆向解决问题的新思路,即从正常的光谱入手,根据正常光谱的特征先确定哪些像素是属于正常光谱的,然后类似排除法地筛选出被宇宙射线污染的像素点。2.介绍了模板匹配的原理和基本步骤。并以此为基础,针对LAMOST拍摄所得的二维光纤光谱图像的特殊情况,对原有基本算法做了部分的修改。利用改良后的模板匹配算法,实现了之前提出的逆向检测算法。实验结果证明了其优异的检测性能。3.在原有的算法基础上,引入现今流行的基于GPU的通用计算的概念。通过GPU来进行匹配运算,克服了模板匹配耗时较长的弊病。对比实验展示了GPU强大的计算能力,并验证了改进的可行性和有效性。4.作为模板匹配的一个衍生,利用模板来有效地修复被宇宙射线污染的像素。并可以利用三维重建中常用的RBF插值算法作为参考,改善了修复的效果。更多还原

【Abstract】 Large Sky Area Multi-Object Fiber Spectroscopy Telescope (LAMOST) is located at Xinglong Observing Station of National Astronomical Observatories, Chinese Academy of Sciences. As a large fiber spectroscopy telescope, it can possess the highest spectrum acquiring rate in the world. LAMOST is a milestone in the history of Chinese astronomical researching.In nature, LAMOST is a complex set of imaging system. Hence, plenty of image processing algorithms are suitable here to be used to process the spectroscopic CCD images. There are many problems may occurred during observing. All these should be processed properly after that. This paper will focus on the inevitable problem of cosmic-ray hits in spectroscopic CCD images.After introducing the spectroscopic CCD images and cosmic-ray hits in detail, we propose a novel method to reject and restore cosmic-ray hits based on template matching algorithm. Then the idea of general purpose computation on graphic hardware is applied to speed up our algorithm. Radial basis function interpolation is also introduced here to achieve better restoring results. In summary, the main works are described as follows,1. Based on the different profile between spectrum and cosmic-ray hits, we provide a new idea to identify cosmic-ray hits. Instead of searching for cosmic-ray hits directly, we confirm those pixels belonged to normal spectrum first. Then the left pixels are affected by cosmic-ray hits.2. Using the Improved template matching, we introduce the procedure of rejecting cosmic-ray hits step by step. The experiments show that our method has impressive performs.3. By introducing the popular idea of GPU computing, we use GPU to do the work of matching. Final result shows that the cost time is shorted greatly.4. As the extension of template matching, templates are used to repair the pixels affected by cosmic-ray hits. With the help of RBF interpolation, we can obtain better restoring results.更多还原