基于数学形态学的LMT信号去噪研究

发布时间：2018-04-28 02:56

本文选题：去噪 + 数学形态学　；参考：《成都理工大学》2017年硕士论文

【摘要】：自从大地电磁测深法问世以来,因其方法简单、实现便利、效果显著等诸多优点在越来越多的领域得到应用。而随之而来的大地电磁信噪分离方法研究也成为信号分析领域的一个热点和难点问题,特别长周期大地电磁信号因其信号能量弱,频域宽,容易被干扰等因素,成为信号的信噪分离的一个难点问题。而受干扰的大地电磁信号将会对后续资料解释产生严重影响,所以在野外采集的长周期大地电磁信号都需要先经过去噪处理后才能更可靠的使用。目前针对长周期大地电磁信号的去噪虽然方法有很多,诸如小波变换、加窗傅里叶变换、源参考法等,也取得了许多研究成果,但目前的方法都有一定的缺陷,且大部分都是基于频域去噪提出来的。因此提出更多更适应长周期大地电磁信号的处理方法一直是大地电磁测深领域内的热点,本文基于此背景,提出了数学形态学原理对长周期大地电磁信号进行处理。本文提出的数学形态学方法与过去对长周期大地电磁信号处理方法的不同之处是数学形态学去噪是基于时域进行处理,适合大地电磁信号频域宽的特点,而且此方法数学理论基础扎实,原理简单,发展空间大。本文的仿真实验也表明此方法是十分有效的。本文主要内容如下:(1)总结了大地电磁信号的场源,分析了大地电磁的各类噪声干扰。并分析每种噪声干扰对大地电磁信号造成的不同影响。指出长周期大地电磁数据存在数据确实现象。(2)对信号的分类和信号的一般处理方法进行了简要的介绍。对不同信号的性质也做了简明扼要的说明,介绍了信号的各种分析方法中的典型算法。同时介绍的信号频谱,幅度等特征(3)介绍了形态学的基本数学理论,详细阐述了形态学滤波中的四种基本算子的定义与数学含义。并用实验仿真了四种基本算子的滤波效果,讨论了它们具体的去噪性能。然后对分析了形态学滤波中的传统的结构元素类型,用几何和二值化参数对它们进行了解释。模拟和对比了传统几种结构元素的去噪性能,并自定义结构元素来对信号进行处理。(4)在数学形态学滤波的理论基础上对长周期大地电磁实测信号做了去噪处理,分析了不同类型结构元素对长周期大地电磁的去噪效果,分析了数学形态学的数据数据填补功能。并在前面的基础上探讨了改进型的数学形态学滤波器,用实测信号对其进行分析。在基于数学形态学去噪的基础上结合中值滤波对长周期大地电磁信号进行处理,实验结果表明这种方法也是有效的。(5)最后用本文提出的数学形态学的处理方法与目前去噪领域中常用的小波包去去噪方法同时对同一信号进行处理,实验结果也表明在此信号的处理上本文提出的方法略优于小波包去噪方法。
[Abstract]:Since the advent of magnetotelluric sounding (MT), it has been applied in more and more fields because of its simplicity, convenience and remarkable effect. The research of magnetotelluric signal noise separation has become a hot and difficult problem in the field of signal analysis, especially because of its weak signal energy, wide frequency domain and easy to be interfered. It is a difficult problem for the separation of signal and noise. The disturbed magnetotelluric signal will have a serious impact on the interpretation of the subsequent data, so the long period magnetotelluric signals collected in the field need to be de-noised before they can be used more reliably. Although there are many methods of de-noising for long-period magnetotelluric signals, such as wavelet transform, windowed Fourier transform, source reference method and so on, many achievements have been made, but the present methods have some defects. And most of them are based on frequency domain denoising. Therefore, it has been a hot spot in the field of magnetotelluric sounding to put forward more and more suitable processing methods for longperiod magnetotelluric signals. Based on this background, this paper puts forward the mathematical morphology principle to process long-period magnetotelluric signals. The difference between the mathematical morphology method proposed in this paper and the previous methods of long period magnetotelluric signal processing is that the mathematical morphology denoising is based on time domain processing, which is suitable for the characteristics of magnetotelluric signal frequency domain width. Moreover, this method has a solid foundation of mathematics theory, simple principle and large space for development. The simulation results also show that this method is very effective. The main contents of this paper are as follows: (1) the field source of magnetotelluric signal is summarized, and the noise interference of magnetotelluric signal is analyzed. The different effects of each noise disturbance on magnetotelluric signal are analyzed. It is pointed out that long period magnetotelluric data exist in real data. (2) the classification of signals and the general processing methods of signals are briefly introduced. The properties of different signals are also explained concisely, and the typical algorithms in various signal analysis methods are introduced. The basic mathematical theory of morphology is introduced, and the definition and mathematical meaning of four basic operators in morphological filtering are described in detail. The filtering effect of four basic operators is simulated by experiments, and their specific denoising performance is discussed. Then, the traditional structural elements in morphological filtering are analyzed, and the geometric and binary parameters are used to explain them. This paper simulates and compares the de-noising performance of several traditional structural elements, and defines the structural elements to process the signal. (4) on the basis of mathematical morphological filtering theory, the long-period magnetotelluric signal is de-noised. The de-noising effect of different structural elements to long-period magnetotelluric (LPM) is analyzed, and the data filling function of mathematical morphology is analyzed. On the basis of the above, the improved mathematical morphological filter is discussed, and the measured signal is used to analyze the filter. The long period magnetotelluric signal is processed based on mathematical morphology denoising and median filtering. The experimental results show that this method is also effective. Finally, the same signal is processed simultaneously by the mathematical morphology method proposed in this paper and the wavelet packet de-noising method commonly used in the field of denoising. The experimental results also show that the proposed method is slightly better than the wavelet packet denoising method in the signal processing.
【学位授予单位】：成都理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN911.7

【参考文献】