基于分数阶小波变换的地震数据随机噪声压制方法研究

发布时间：2018-02-22 16:49

本文关键词： 地震数据随机噪声分数阶B样条小波高斯尺度混合模型　出处：《合肥工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：地震数据中的随机噪声是一种干扰波,它严重降低了地震数据的质量和信噪比,并影响着资料的后续处理和分析。因此,对于受干扰的地震数据进行的降噪处理工作显得尤为重要。地震数据降噪就是指根据野外采集到的实际地震数据估计恢复出原始地震数据,即在削减噪声的同时较好的保留实际地震数据中的有效信息。分数阶B样条小波具有可选的阶次,可以更灵活地调节小波系数;高斯尺度混合模型能够很好地描述小波变换系数的边缘分布和邻域系数之间的相关性。本文将分数阶B样条小波变换与高斯尺度混合模型相结合用于压制地震数据中的随机噪声。本文的主要研究工作为:1.综述了地震数据降噪的发展,阐述了分数阶B样条小波的构成和性质,以及高斯尺度混合模型,介绍了地震数据降噪的评价标准。2.将分数阶B样条小波变换与高斯尺度混合模型结合用于地震数据随机噪声的压制。利用分数阶B样条小波变换将含噪地震数据映射到最优分数阶小波时频域内,然后对各子带小波系数分别建立高斯尺度混合模型,由贝叶斯方法估计出源地震数据小波系数,最后使用分数阶B样条小波逆变换重构得到降噪后的地震数据。通过对合成地震记录和实际地震数据进行降噪处理,表明该方法能够有效地压制地震数据中的随机噪声,较好地保留了有效信号。
[Abstract]:The random noise in seismic data is a kind of interference wave, it reduces the seismic data quality and signal-to-noise ratio, and affect the subsequent processing and analysis of data. Therefore, noise reduction for interference of seismic data is particularly important. It refers to the seismic data denoising according to the actual seismic data acquisition to estimate the recovery of the original seismic data, namely in the reduction of noise and effective information retention of real seismic data well. The fractional order B spline wavelet is optional, can adjust flexibly the wavelet coefficients; Gauss scale mixture model can well describe the relationship between the marginal distribution and neighborhood coefficients of wavelet transform coefficient. In this paper, the fractional order B spline wavelet transform and the Gauss scale hybrid model combining used to suppress random noise in seismic data. The main research works are as follows: 1. summarizes the development of seismic data denoising, describes the composition and properties of fractional order B spline wavelet, and the Gauss scale mixture model, introduces the evaluation criteria of.2. seismic data denoising and the fractional order B spline wavelet transform and Gauss scale hybrid model combined with the seismic data used to suppress random noise. Using fractional B like a wavelet transform denoising of seismic data is mapped to the optimal fractional wavelet in time-frequency domain, and wavelet coefficients of each scale are established by the Gauss mixture model, Bayesian method to estimate the source seismic data of seismic data using wavelet coefficients, finally B fractional spline wavelet transform reconstruction are denoised. Through the synthesis seismic record and real seismic data denoising, show that this method can effectively suppress random noise in seismic data, retaining the effective signal.

【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P631.4

【相似文献】