基于Radon变换的时频峰值滤波在地震资料去噪上的应用

发布时间：2018-05-06 05:06

本文选题：高分辨率Radon变换 + 时频峰值滤波(TFPF)　；参考：《吉林大学》2015年硕士论文

【摘要】：地震勘探是探索地震结构，判断油层储气有无及位置的重要方法。这种方法主要通过高能冲击作为大地的输入，通过检波器收集响应信号，从而判断地质界面的属性及地质结构。但是由于油气资源的过度利用，使得勘探者很难在浅表位置找到油气资源，地震勘探技术就要向更深的地域发展。由于信号传播距离远，信号的幅值衰减严重，并会被随机噪声淹没，这将导致无法识别信号。因而地震去噪就尤为重要了。时频峰值滤波作为新兴的去噪方法在地震噪声去除上获得很好的效果，近些年来引起了地球物理届的关注。但传统时频峰值滤波仅沿时间方向进行滤波，没有利用信号的时空域相关性恢复信号，这将影响信号的保留。其次传统算法对地震记录采用固定的窗长滤波，大窗长可以很好的压制噪声，但信号保幅不理想；小窗长信号保幅理想，但噪声仍然大量遗留。为达到时频峰值滤波保幅性能和压噪性能兼顾的目标，本论文提出了新的二维时空域滤波方法，即高分辨率双曲Radon域变窗长时频峰值滤波，这种算法首先利用高分辨率Radon变换将双曲同相轴集中于Radon域下的几个高能量位置。由于有效信号的叠加，在Radon域信号部分更容易识别，这为后续的滤波提供了有利条件。之后在Radon域通过设定合理的阈值，识别出信号和噪声，并对信号采用小窗长滤波，，对噪声采用大窗长滤波，这样可以做到保幅的同时，去掉更多的随机噪声。同时，为减弱时频峰值滤波对高频信号的衰减，本文提出了一种新的算法，即变曲率双曲轨线时频峰值滤波。该算法首先利用Radon变换把地震记录不同弯曲程度的同相轴分离不同的子记录，之后对每一个子记录选择合适的采样轨线，使得采样后的频率降低，这样经过TFPF后信号幅度衰减将减小此方法可以减少TFPF对有效信号带来的误差，从而提高滤波结果的信噪比。我们将此算法应用到了模拟以及实际地震记录中，都达到了很好的保幅和去噪的结果。
[Abstract]:Seismic exploration is an important method to explore seismic structure and determine the location of reservoir gas. This method mainly uses high-energy shock as the input of the earth and collects the response signal by geophone to judge the properties and geological structure of the geological interface. However, due to the overuse of oil and gas resources, it is difficult for prospectors to find oil and gas resources in shallow positions, and seismic exploration technology will develop to deeper areas. Because the signal propagates far away, the amplitude of the signal attenuates seriously, and will be submerged by random noise, which will lead to the failure to recognize the signal. So seismic denoising is particularly important. As a new denoising method, time-frequency peak filtering has achieved good results in seismic noise removal, which has attracted the attention of geophysics in recent years. However, the traditional time-frequency peak filter only filters along the time direction and does not recover the signal by using the temporal and spatial correlation of the signal, which will affect the retention of the signal. Secondly, the traditional algorithm uses the fixed window length filter to the seismic records, the large window length can suppress the noise very well, but the signal amplitude preservation is not ideal; the small window length signal keeps the amplitude ideal, but the noise is still largely left behind. In order to achieve both amplitude preserving performance and denoising performance of time-frequency peak filtering, a new two-dimensional spatio-temporal filtering method is proposed in this paper, that is, high-resolution hyperbolic Radon domain variable window time-frequency peak filtering. In this algorithm, the hyperbolic cophase axis is firstly concentrated in several high energy positions in Radon domain using high-resolution Radon transform. Because of the superposition of the effective signal, it is easier to identify the signal in the Radon domain, which provides favorable conditions for the subsequent filtering. After that, the signal and noise are identified in Radon domain by setting a reasonable threshold, and the signal is filtered with small window length, and the noise is filtered with large window length, which can preserve the amplitude and remove more random noise at the same time. At the same time, in order to reduce the attenuation of high frequency signals by time-frequency peak filtering, a new algorithm, time-frequency peak filtering for hyperbolic track with variable curvature, is proposed in this paper. The algorithm firstly uses Radon transform to separate different sub-records from the same phase axis of seismic records with different bending degrees, and then selects appropriate sampling tracks for each sub-record, so that the frequency after sampling is reduced. In this way, the signal amplitude attenuation after TFPF can reduce the error caused by TFPF to the effective signal, thus improving the signal-to-noise ratio of the filtering results. The algorithm is applied to the simulation and the actual seismic records, and the results of amplitude preserving and denoising are very good.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P631.4;TN911.7

【参考文献】