太阳高分辨图像的多尺度速度场测量方法研究

发布时间：2018-06-15 15:03

本文选题：速度场 + 局部相关跟踪　；参考：《昆明理工大学》2017年硕士论文

【摘要】：太阳图像中蕴含着大量不同尺度、不同频率和不同速度的太阳活动的运动学信息,比如大尺度的活动(黑子、耀斑、暗条等)以及小尺度活动(微耀斑、微暗条等)。当不同尺度的太阳活动混合在一起的时候,特定目标的速度场的测量就变得非常困难,小尺度的变化会被大尺度的运动所掩盖。本文的研究目标是利用多尺度多分辨率的思想与光流技术相结合,计算不同空间尺度的速度场,以进行混合运动速度场的测量。从而得到各种尺度太阳活动特征的运动趋势,全面地从局部到整体理解太阳特征的运动规律,为太阳物理学家提供研究太阳活动演化的有效工具。本文主要进行了以下研究:1.论文分析了传统的单一尺度速度场测量方法,然后对局部相关跟踪算法进行改进,称为加权局部相关跟踪算法。改进的内容主要为粗细匹配的搜索策略、将最大互相关系数作为权重对局部区域进行加权平均。并对局部相关跟踪和加权局部相关跟踪这两种算法进行了比较;2.为了测量不同尺度的太阳活动的速度场,将加权局部相关跟踪算法与多尺度多分辨率分层细化的思想相结合,实现了多尺度加权局部相关跟踪算法,以测量混合运动速度场;3.为了验证多尺度加权局部相关跟踪算法的精度,首先通过模拟实验进行误差分析。分析结果表明我们所改进的多尺度加权局部相关跟踪算法的测量误差为1%;然后对三组真实的数据集进行速度场的测量。不仅对黑子本影和半影的速度场进行准确测量,而且在其中的两组数据集中,还发现黑子本影内有运动,这是传统的单一尺度速度场测量方法不能得到的;4.讨论了多尺度加权局部相关跟踪算法的重要参数,包括金字塔的层数、计算窗口大小的设置。本文所提出的方法不仅仅可用来分析相邻时刻的变化,更重要的是从整个序列图像的角度去考虑图像内容的变化。通过模拟实验以及真实的数据测量,我们发现该方法不仅能够精确测量单一尺度运动的速度场,而且可以解决混合运动中大尺度淹没小尺度运动的问题。通过构建金字塔图像,可以定量的分析不同分辨率下的速度场,得到在不同空间尺度的速度场,从而体现各种尺度太阳活动特征的运动趋势。
[Abstract]:The solar images contain a great deal of kinematic information of solar activities with different scales, frequencies and velocities, such as large scale activities (sunspots, flares, dark strips, etc.) and small scale activities (micro flares, small dark strips, etc.). When the solar activities of different scales are mixed together, the measurement of the velocity field of a specific target becomes very difficult, and the small scale changes will be masked by the large scale motion. The purpose of this paper is to calculate the velocity field of different spatial scales by combining the idea of multi-scale and multi-resolution with optical flow technique to measure the velocity field of mixed motion. Thus, the motion trend of solar activity characteristics on various scales is obtained, and the motion law of solar characteristics is comprehensively understood from local to global, which provides an effective tool for solar physicists to study the evolution of solar activity. This article mainly carries on the following research: 1. This paper analyzes the traditional single scale velocity field measurement method, and then improves the local correlation tracking algorithm, which is called weighted local correlation tracking algorithm. The main content of the improvement is the search strategy of coarse and fine matching, in which the maximum cross-relation number is used as the weight to weighted the local region. The local correlation tracking and weighted local correlation tracking are compared. In order to measure the velocity field of solar activity at different scales, the weighted local correlation tracking algorithm is combined with the idea of multi-scale multi-resolution hierarchical thinning, and a multi-scale weighted local correlation tracking algorithm is implemented to measure the velocity field of mixed motion. In order to verify the accuracy of the multi-scale weighted local correlation tracking algorithm, error analysis is carried out through simulation experiments. The analysis results show that the measurement error of our improved multi-scale weighted local correlation tracking algorithm is 1 and then the velocity field of three groups of real data sets is measured. Not only the velocity fields of the sunspot shadow and penumbra are measured accurately, but also in the two sets of data sets, it is found that there is movement in the shadow of the sunspot, which can not be obtained by the traditional method of single scale velocity field measurement. The important parameters of the multi-scale weighted local correlation tracking algorithm are discussed, including the number of layers of the pyramid and the setting of calculating window size. The method proposed in this paper can be used not only to analyze the changes of adjacent moments, but also to consider the change of image content from the point of view of the whole sequence of images. Through simulation experiments and real data measurements, we find that this method can not only accurately measure the velocity field of a single scale motion, but also solve the problem of submergence of large scale and small scale motion in mixed motion. By constructing pyramid image, the velocity field at different resolution can be quantitatively analyzed, and the velocity field at different spatial scales can be obtained.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P182;TP391.41

【参考文献】