【作者】 潘洪屏；

【导师】 刘霞；

【作者基本信息】 东北石油大学 ， 模式识别与智能系统， 2011， 硕士

【摘要】 地震勘探是一种重要的物探方法,野外采集的地震资料中包含着有关地下构造和岩性的信息,但这些信息与干扰背景相叠加,并且被外界因素所扭曲,不宜直接利用野外资料做地质解释,因此提高其信噪比是地震信号处理的重要任务。本论文着重研究了基于小波分析技术的地震勘探信号的去噪方法。本文首先介绍地震勘探信号中噪声的特点及形成原因,然后详述小波分析理论;接着研究小波阈值去噪、小波变换模极大值去噪和小波包分析三种去噪方法。由于小波阈值去噪算法中阈值是通过经验公式选取的,对信噪比较低的地震信号去噪效果不好,有效信号损失较大,分辨率低,针对常用阈值选取的这些不足,提出了基于小波熵高分辨率阈值去噪方法,先经过相关性处理保留有效信号的高频信息,提高分辨率,再根据小波熵来选取阈值,小波熵是小波变换和信息熵的结合,通过对含噪的ricker子波、合成地震信号和实际地震信号去噪,验证了该改进算法具有很好的去噪效果。小波变换模极大值去噪算法中最大尺度上的阈值通常根据经验选取,并且在搜索由有效信号产生的模极大值的传播点时容易出现错选现象,针对小波变换模极大值去噪算法的这些不足,提出了小波熵和相关性结合的模极大值去噪方法,根据小波熵来选取最大尺度上的阈值,利用相关性来解决在搜索过程中出现的错选问题,通过对含噪的ricker子波、合成地震信号和实际地震信号去噪,验证了该改进算法具有很好的去噪效果。另外,还运用小波包分层阈值和全局阈值对含噪的ricker子波、合成地震信号和实际地震信号进行去噪处理,小波包分层阈值去噪也具有很好的去噪效果。更多还原

【Abstract】 Seismic exploration is an important geophysical method, seismic data collected in the field includes the information about the subsurface structure and lithology, but the information is overlaid by background interference, and distorted by external factors. So field datas can not be used to do geological interpretation directly. Thereby, increasing signal to noise ratio is an important task of seismic signal processing. This paper focuses on research of seismic signal de-noising ways based on wavelet analysis.This paper firstly introduces the characteristics and causes of the noise included in the seismic signal, and then describes wavelet analysis theory in detail; then studies wavelet threshold de-noising, wavelet transform modulus maxima de-noising and wavelet packet analysis. The threshold in the wavelet threshold de-noising algorithm is selected by empirical formula. The de-noising effect is not good for seismic signal of low signal to noise ratio, because the results lead to large effective signal loss and low resolution. For these deficiencies of common threshold selection, we propose a new de-noising method based on wavelet entropy high-resolution threshold. In order to reserving the high frequency information of the effective signal and improving resolution, the high-frequency wavelet coefficient is processed by the correlation. Then the threshold is selected by the wavelet entropy. The wavelet entropy is the combination of wavelet transform entropy and information entropy. Through the de-noising of noisy ricker wavelet, synthetic seismic signal and actual seismic signal, the good effect of this algorithm has been proved. The largest scale threshold in wavelet transforms modulus maxima de-noising algorithm is selected by Exp, and it is prone to be a wrong choice during the process of searching the modulus maximum points generated by valid signal. For these deficiencies of wavelet transform modulus maxima de-noising, we propose a wavelet entropy and correlation with the modulus maxima de-noising method. The threshold on the largest scale is selected by the wavelet entropy, and the wrong choice issue during the the process of searching is solved by the correlation. The improved algorithm has a good de-noising effect during the the process of de-noising the noisy ricker wavelet, the synthetic seismic signal and the actual seismic signal. In addition, we use the method of wavelet packet layered threshold and global threshold to process the noisy ricker wavelet, the synthetic seismic signal and the actual seismic signal. The wavelet packet layered threshold also has a good de-noising effect.更多还原