基于NVST观测的米粒识别与形态特征分析的研究

发布时间：2017-12-27 17:40

本文关键词：基于NVST观测的米粒识别与形态特征分析的研究　出处：《昆明理工大学》2016年硕士论文　论文类型：学位论文

【摘要】：米粒组织是太阳表层对流运动所产生的一种形态特征。研究米粒的形态特征和演化过程能帮助我们更好地了解太阳表层对流运动的产生机制以及这种机制和太阳磁场活动之间的关系。因此,寻找一个精准的识别方法对于研究这种机制和关系是十分迫切的。然而,由于在米粒图像中米粒组织的边缘对比度低、模糊等问题,使得基于强度或梯度的米粒特征识别方法面临很大的困难。为了解决上述问题,本文运用相位一致性技术来对米粒结构特征进行识别和提取。本文选用中国科学院云南天文台抚仙湖的新一代太阳真空望远镜(New Vaccum Sloar Telescope,简称NVST)的2组高分辨活动区观测米粒组织图像来阐述算法的识别过程。米粒的识别过程包含三步：(1)获得相位一致信息图；(2)获得二值图；(3)通过形态学滤波提取米粒结构。本文选取了3个目前已存在的米粒识别算法(阂值技术、标记分水岭和高斯拉普拉斯算子)同本文的识别方法做了对比实验目的是验证所提算法的准确性和有效性。实验结果表明所提算法能够有效地提取那些低对比度的米粒特征。同时为了检验算法对阈值的响应程度,分别选取了3组阈值来检验不同阈值情况下的差异性。结果证明所提算法对阈值响应不敏感。利用本文所提方法,数据A和数据B分别识别出165694和108279米粒结构。首先获得每个米粒的直径,根据米粒直径的分布情况将米粒分为了两类：直径大于780km的regular米粒和小于780km的mini米粒。并且根据分类结果分析了米粒的其他几个特征：米粒的直径、强度、形状以及分形维数进行统计。其目的就是进一步验证提算法获得统计结果的准确性,从统计结果来看,米粒的特征统计结果很好地符合了已有文献的结论,这也进一步验证所提算法的准确性与鲁棒性,能够用于进一步的米粒演化以及物理机制的科学研究。
[Abstract]:Grain tissue is a morphological characteristic of the convective movement of the surface of the sun. Studying the morphological characteristics and evolution process of rice grains can help us to better understand the mechanism of solar surface convection and the relationship between this mechanism and solar magnetic field activities. Therefore, it is very urgent to find a precise method of recognition for the study of this mechanism and relationship. However, because of the low contrast and blurred edge of rice grain in rice grain image, the feature recognition method based on intensity or gradient is very difficult. In order to solve the above problems, this paper uses phase consistency technology to identify and extract the structure characteristics of rice grain. A new generation of solar vacuum telescope the China Academy at Yunnan Observatory in the Fuxian Lake (New Vaccum Sloar Telescope, referred to as NVST) of the 2 groups of high resolution observation activities granulation images to illustrate the recognition algorithm. The recognition process of rice grain consists of three steps: (1) obtaining the phase consistent information map; (2) obtaining the two value diagram; (3) the grain structure is extracted by morphological filtering. This paper selects 3 existing rice identification algorithm (threshold technology, Gauss Laplasse and mark watershed operator) with the identification method for experiment is designed to validate the proposed algorithm is accurate and effective. The experimental results show that the proposed algorithm can effectively extract the characteristics of rice grains with low contrast. At the same time, in order to test the degree of response of the algorithm to the threshold, 3 sets of thresholds are selected to test the difference between different threshold conditions. The results show that the proposed algorithm is not sensitive to the threshold response. Using the method proposed in this paper, data A and data B identify the structure of 165694 and 108279 meters respectively. First, the diameter of each grain was obtained. According to the distribution of the grain diameter, the rice grains were divided into two groups: the regular grains larger than 780km and the mini grains less than 780km. The other characteristics of rice grain are analyzed according to the classification results: the diameter, strength, shape and fractal dimension of rice grain are counted. Its purpose is to further verify the accuracy of the proposed algorithm to obtain statistical results, statistically, the characteristics of the statistical results of rice is in good agreement with the results of previous literature, which further verified the accuracy and robustness of the algorithm, can be used for further scientific research and the evolution of the physical mechanism of rice.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：P111.41;P182.2

【相似文献】