基于卫星高度计的中尺度涡自动识别算法研究

发布时间：2018-07-10 19:07

本文选题：中尺度涡 + WA算法　；参考：《国家海洋环境预报中心》2015年硕士论文

【摘要】：中尺度涡是普遍存在于世界各大洋的一种中尺度海洋现象。近年来,随着卫星高度计的迅速发展,采用大覆盖、准同步、长时间连续观测的高度计资料研究全球或特定海域中尺度涡的统计特征和传播特征已成为当今的一大研究热点。因此,如何运用自动识别技术在这海量的卫星数据中快速、准确的识别中尺度涡就显得尤为重要,这也成为定量计算中尺度涡的各项统计特征的基础。本文主要针对目前为止使用最广泛的3种涡旋自动识别算法——WA算法、OW算法和无阈值等值线法,在实现并总结原算法的基础上,对各算法所存在的问题分别加以改进,以提高算法的识别精度和计算效率。同时对三种改进算法进行对比分析,采用识别精度最高的WA算法与CTD实测资料相结合,探讨了2013年夏秋两季越南东南外海暖涡的演化特征、水文特征和内部结构。具体工作如下:1)本文在介绍了WA算法实现过程中用到的处理技巧和心得体会的基础上,针对WA算法计算量大、计算过程复杂的问题,提出了一种新的“流线聚类”算法以及与之对应的“涡旋边缘识别算法”。新的流线聚类算法不仅计算量较小,而且无需人为设定聚类半径阈值,精度更高、聚类结果更加可靠;两种新算法的结合也解决了原算法无法识别内部结构较为复杂的多涡结构的问题。2)本文针对OW算法常见的两个问题:受卫星高度计数据噪声影响显著、同一W等值线内常包含极性相反的涡旋,分别进行了改进。尤其针对后一个问题,提出了两种分离算法,即“涡度法”和“单连通区域法”,并对它们各自的实验结果进行了对比和分析。3)本文利用SLA数据对无阈值等值线法进行检验,检验结果证实该算法不仅适用于SSH数据,也很好的适用于SLA数据。在此基础上我们针对原算法无法识别涡核位置以及多涡结构的问题,提出了一种“边核匹配”算法,该算法能够准确识别涡旋中涡核的数量、每个涡核的位置以及强度等。4)本文在三种改进算法之间进行比较分析,通过专家目视判别的方式定量地统计每一种算法的识别精度,通过计算每种算法的成功识别率、过量识别率、漏识率、类信噪比、面积重叠率等,定量分析各算法间的精度差异,并考虑到了由专家主观因素引起的评估结果的不确定性,给出了允许的误差上限。5)本文采用识别精度最高的WA涡旋自动识别算法,对2013年夏秋两季越南东南外海暖涡的演化特征做了初步分析,并结合CTD实测资料对比验证WA算法的识别结果,分析越南东南外海暖涡在演化过程中内部性质、结构发生的变化。
[Abstract]:Mesoscale vortex is a kind of mesoscale ocean phenomenon that exists in every ocean of the world. In recent years, with the rapid development of satellite altimeters, the use of large coverage, quasi-synchronous, long-term continuous observation altimeter data to study the global or specific sea area mesoscale vortex statistical characteristics and propagation characteristics has become a major research hotspot. Therefore, how to use automatic recognition technology to identify mesoscale vortices quickly and accurately in these massive satellite data is particularly important, which becomes the basis of quantitative calculation of the statistical features of mesoscale vortices. This paper mainly aims at three kinds of vortex automatic recognition algorithms, WA algorithm OW algorithm and non-threshold contour method, based on the realization and summary of the original algorithm, the existing problems of each algorithm are improved respectively. In order to improve the recognition accuracy and computational efficiency of the algorithm. At the same time, three improved algorithms are compared and analyzed. The evolution, hydrological characteristics and internal structure of warm vortex in Southeast Vietnam Sea in summer and autumn 2013 are discussed by combining WA algorithm with CTD data. The specific work is as follows: 1) on the basis of introducing the processing skills and experience used in the implementation of WA algorithm, this paper aims at the problems of the large amount of calculation and complex calculation process of WA algorithm. A new "streamline clustering" algorithm and a corresponding "vortex edge recognition algorithm" are proposed. The new streamline clustering algorithm not only needs less computation, but also does not need to set the threshold of clustering radius artificially, so the accuracy is higher and the clustering result is more reliable. The combination of two new algorithms also solves the problem that the original algorithm can not identify the complex multi-vortex structure of the internal structure. This paper aims at two common problems of the OW algorithm: it is significantly affected by the satellite altimeter data noise. The vortex with opposite polarity is often contained in the same W isoline, which is improved respectively. Especially for the latter problem, two separation algorithms, vorticity method and single connected region method, are proposed, and their respective experimental results are compared and analyzed. 3) in this paper, SLA data are used to test the contours without threshold. The results show that the algorithm is not only suitable for SSH data, but also suitable for SLA data. On the basis of this, we propose a "edge kernel matching" algorithm, which can accurately identify the number of vortex cores in a vortex, aiming at the problem that the original algorithm is unable to identify the location of vortex cores and the structure of multiple vortices. The location and intensity of each vortex core are compared and analyzed among the three improved algorithms. The recognition accuracy of each algorithm is quantitatively counted by the method of expert visual discrimination, and the recognition rate of each algorithm is calculated. Excess recognition rate, miss recognition rate, class signal-to-noise ratio, area overlap rate, etc., quantitative analysis of the accuracy differences among the algorithms, and taking into account the uncertainty of the evaluation results caused by the subjective factors of experts, The allowable upper limit of error is given. (5) in this paper, the evolution characteristics of warm vortex in southeast Vietnam in summer and autumn 2013 are preliminarily analyzed by using WA vortex automatic recognition algorithm with the highest recognition accuracy. The results of WA algorithm are compared with CTD data, and the internal properties and structure changes of warm vortices in southeast Vietnam are analyzed.
【学位授予单位】：国家海洋环境预报中心
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P715.6

【参考文献】