基于智能学习的X波段海浪信息参数反演算法

发布时间：2018-06-04 19:14

本文选题：X波段雷达 + 海浪参数　；参考：《电子科技大学》2017年硕士论文

【摘要】：21世纪是人类海洋开发利用的新时代,对海洋资源的开发利用是人类经济进一步发展的重要方向。海洋对于一个国家的国防安全至关重要,特别是近些年来我国海洋权益不断遭受挑衅。同时,海洋也在不断经受人类的各种污染,海洋环境日益恶化。对于海洋的了解和研究一直是各种学科领域的热点,特别是对海洋物理环境的监测,关系到诸多关键领域,对于社会经济的发展意义重大。总之,只有更好的研究和了解海洋,海洋才能提供给我们无尽的资源。X波段雷达,具有时间空间分辨率高,成本低廉,工作稳定,安装方便等优点。人们使用该设备进行海洋探测预警已有很长时间,由于其自动运行,无需值守,能够24小时全天候工作,X波段雷达对于浪流监测也非常方便快捷。本论文,首先回顾了已经成熟的几种海面浪流参数监测方法,以及这些年来利用X波段雷达进行海面参数监测的发展现状。了解了雷达在海面成像以及调制原理后,初步介绍了人工神经网络的概念以及发展现状。之后,针对海面表层流参数的反演,分别采用了最小二乘算法(Least Squares Method)、迭代最小二乘算法(Iterative LSM)、NSP(Normalized Scalar Product)算法。现实中海浪并不是理想的线性谱叠加的关系,特别是在复杂海况下,根据线性关系得到的反演算法并不能非常精确的表征雷达图像与海面浪流参数的关系。我们在计算有效波高时使用的调制传递函数为线性拟合的关系,所以计算精度有限。因此,针对有效波高,本文搭建调试一个广义回归神经网络来进行海谱的拟合以及参数的反演,建立雷达图像一维频谱与浮标谱之间的映射关系。海面雷达图像的数值模拟,可以帮助我们了解后向散射信号形成机理,同时对模拟图像的计算可以验证三种算法的优劣。在阐述了海浪谱的表达形式以及拟合雷达谱与模拟海谱后。针对不同算法进行表层流以及海浪参数的模拟反演,证明了相应算法的准确可靠。利用已经测量得到的大量海面雷达图像数据以及对应浮标海流计数据,本文最后进行了实测雷达数据的参数反演,得到结果后与相对应的浮标或者海流计数据进行对比分析。针对有效波高的反演,分别采用传统线性拟合参数定标的算法和广义回归神经网络的算法进行横向对比。经计算分析,各种算法对于相应的海面浪流参数反演结果都比较好,有效波高的计算结果表明,神经网络算法能够取得更好的结果。对比浮标和海流计参数,各参数计算结果相关系数均在0.75之上。
[Abstract]:The 21st century is a new era for the exploitation and utilization of human ocean. The exploitation and utilization of marine resources is an important direction for the further development of human economy. The ocean is very important for the national defense security of a country, especially in recent years, the maritime rights and interests of our country have been constantly challenged. At the same time, the ocean is also suffering from various kinds of human pollution, the marine environment is deteriorating day by day. The understanding and research of the ocean has always been a hot spot in various disciplines, especially the monitoring of the marine physical environment, which relates to many key fields and is of great significance to the development of social economy. In a word, only by better studying and understanding of the ocean can the ocean provide us with endless resources. X-band radar has the advantages of high spatial and temporal resolution, low cost, stable work and convenient installation. It has been used for a long time for ocean detection and early warning. Because of its automatic operation and no need to be on duty, X-band radar can work 24 hours a day. It is also very convenient and quick to monitor waves and currents. In this paper, we first review several methods of ocean surface current parameters monitoring, and the development status of using X-band radar to monitor sea surface parameters in recent years. After understanding the principle of radar imaging and modulation at sea level, the concept and development of artificial neural network (Ann) are introduced. After that, the least square algorithm and iterative least square algorithm are used to retrieve the surface current parameters, respectively, and the iterative least square algorithm (LSM) is used to calculate the NSPN Normalized Scalar Product (Normalized Scalar Product) algorithm. In reality, ocean wave is not an ideal linear spectrum superposition relation, especially in complex sea conditions, the inversion algorithm based on linear relationship can not represent the relationship between radar image and sea surface current parameters very accurately. The modulation transfer function used in calculating the effective wave height is linear fitting, so the calculation accuracy is limited. Therefore, aiming at the effective wave height, a generalized regression neural network is built to fit the sea spectrum and inverse the parameters, and the mapping relationship between the one-dimensional spectrum and the buoy spectrum of radar image is established. The numerical simulation of sea surface radar image can help us to understand the formation mechanism of backscattering signal. At the same time, the calculation of simulated image can verify the merits and demerits of the three algorithms. The expression of wave spectrum and the fitting of radar spectrum and simulated sea spectrum are described. The simulation and inversion of surface current and ocean wave parameters for different algorithms prove the accuracy and reliability of the corresponding algorithm. Using a large number of sea surface radar image data and the corresponding buoy current data, the parameter inversion of the measured radar data is carried out at the end of this paper, and the results are compared with the corresponding buoy or current meter data. For the inversion of effective wave height, the traditional linear fitting parameter calibration algorithm and the generalized regression neural network algorithm are used for lateral comparison. The results of calculation and analysis show that various algorithms are better for the inversion of sea surface current parameters. The calculation results of effective wave height show that the neural network algorithm can obtain better results. Compared with the parameters of buoy and current meter, the correlation coefficient of each parameter is above 0.75.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN957.52

【参考文献】