基于Radon变换的高密点地震信号去噪方法研究

【作者】 朱大虎；

【导师】 张二华；

【作者基本信息】 南京理工大学 ， 模式识别与智能系统， 2012， 硕士

【摘要】 地震勘探中F-K域滤波被经常使用压制规则干扰,采用多次叠加思想压制随机噪声；但这两种方法在低信噪比和强规则干扰时都具有局限性。效果并不十分理想。对于地震资料干扰,本文主要探讨面波消除。浅震源时,面波会直接产生于大地和空气的分界处。面波的传播速度比横波稍小其频率低,能量在垂直、水平方向上均可衰减,但水平方向上衰减慢,延续时间长。地震剖面上,面波表现为扫帚形并且频散。近偏移距反射波受强烈干扰,能量被面波所湮灭。虽然野外采集使用了压制面波的措施,但采集到的记录中面波干扰仍然存在,影响处理及解释效果。本文探讨了拉东变换的相关知识,对拉东变换进行深入的剖析和研究。使用线性拉东变换、抛物线拉东变换对同一地震数据进行面波消除,结果是线性拉东变换的消除效果明显好于抛物线拉东变换。接着又进行拉东变换中频率域复矩阵算法的改进,经傅氏变换后,所有的矩阵均变换到频率域,若采用的是高斯-约尔当选主元求逆方法,由于程序的时间复杂度为矩阵维数的三次方,对于大矩阵而言,运行时间漫长。本文采用并行算法对高斯-约尔当选主元求逆方法进行改进,时间复杂度变为矩阵维数的平方,取得了较好的效果很好。最后探讨了高密度点地震采集,分析了当前采用高密点采集地震资料的必要性。更多还原

【Abstract】 Seismic exploration in the FK domain filtering rules are frequently used to suppress interference, several times with random noise superimposed thought suppression; but the two methods at low SNR and strong interference rules have limitations. Effect is not very satisfactory. For the seismic disturbance, this paper focuses on the elimination of surface waves.Light source, the surface wave will be directly from the boundary between earth and air. Surface wave propagation velocity smaller than the wave, its frequency is low, the energy in the vertical, horizontal direction can be attenuated, but the slow decay on the horizontal direction, continuing a long time. Seismic profiles, surface wave and the performance of a broom-shaped dispersion. Nearly offset by strong interference from the reflected wave, the wave energy quilt annihilation. Although collected in the field using a surface wave suppression measures, but the records collected in the surface wave interference persists, affect the processing and interpretation of results.This paper discusses the Radon transform of the relevant knowledge, on the Radon transform in-depth analysis and research. Using a linear Radon transform, parabolic Radon transform of the same surface wave seismic data to eliminate, the result is the elimination of the linear effect of the Radon transform is better than the parabolic Radon transform.Followed the Radon transform in the frequency domain complex matrix algorithm, by the Fourier transform, all the matrices are transformed into the frequency domain, the use of the Gauss-Jol was elected principal component inversion method, because of the time complexity is the dimension of the cubic matrix, for large matrices, the long running time. In this paper, parallel algorithms for Gauss-Jol was elected principal component inversion method to improve the time dimension of complexity into the square of the matrix, and achieved good results very well.Finally, we discuss high-density point of seismic acquisition, analysis of the current seismic data acquisition using high-density point of need.更多还原