基于引导滤波和小波变换的太阳冕环结构自动识别算法研究

Automatic detection of coronal loops using guided filtering and wavelet transform

冕环是太阳大气中的炙热等离子在太阳磁场的作用下产生的一种太阳特征结构;研究冕环动力学特征将促进我们对太阳日冕磁场,日冕加热和日冕振荡等研究.冕环结构的准确识别和提取则是相关研究的前提条件.然而由于日冕磁场的复杂性,使得对冕环结构识别和提取面临很大的困境.为此,本文提出了一个新的冕环识别算法.该算法结合了引导滤波和小波变换模极大值的方法来对冕环结构进行自动识别和提取.识别和提取算法过程如下:(1)运用模糊函数对太阳日冕图像的对比度进行增强;(2)使用引导滤波对增强后的图像进行滤波来增强冕环结构的边缘;(3)对滤波过后的图像使用小波变换模极大值法来识别图像中的冕环结构,并对识别出的冕环图像进行二值化处理;(4)对二值化图像进行形态学处理来获取冕环结构.运用该算法对由过渡区和日冕探测器(Transition and Coronal Explorer, TRACE)和太阳动力观测站的大气成像装置(Atmospheric Imaging Assembly on the Solar Dynamics Observatory, SDO/AIA)观测设备在171?波段所观测的日冕图像中的冕环结构进行识别和提取,并同已有文献的识别结果进行了对比,结果表明本文所提出的算法具有很高的识别率和很好的鲁棒性.所识别和提取的冕环结构能进一步用于科学研究中.

The corona is the outer layer of the solar asmosphere.The bright loop-like structures in the corona,which are called coronal loops,are hot plasma to be bound to the solar coronal magnetic fields.Studying these magnetic features can help to better understand the dynamics of coronal magnetic fields and coronal oscillations,and further clearify the debate of coronal heating.Accurate identifying coronal loops are crucial for the relevant studies.However,due to the complexity of the coronal magnetic fields,it is very difficult to identify and extract the loop-like structures.In this case,we propose a novel algorithm based on guided filtering and wavelet transform modulus maxima to automatically detect and extract them.The steps are as follows:we use(1) a fuzzy function to enhance the contrast of solar corona images;(2) guided filtering to highlight the edges of coronal loops;(3) wavelet transform modulus maxima to identitify the edges of coronal loops;(4) image binarization to extract coronal loops;(5) morphological operate to remove those non-loop structures.The images observed by the Transition and Coronal Explorer(TRACE) and the Atmospheric Imaging Assembly on the Solar Dynamics(SDO/AIA) intruments in the 171 ■ channel are selected to illustrate the porcess and further evaluate the performance of the algorithm.The results demonstrate that the proposed algorithm has significant advantages over provious identification methods and the detected and extracted structures can be further applied to scientific researchs.