时空二维方向轨线时频峰值滤波消减地震勘探随机噪声研究

发布时间：2018-08-24 17:11

【摘要】：地震勘探作为地球物理勘探中的一种主要的勘探方法,在资源勘探,地质结构研究中发挥着重要的作用。野外采集的地震资料中往往含有大量的噪声,严重影响了后续的反演和解释工作。因此,对地震资料中的噪声进行消减,提高地震资料的信噪比和分辨率对地质构造的研究以及寻找油气等矿藏资源具有重要意义。国内外现有的消噪方法大多受到某种假设或条件限制,在某些特定条件下,如低信噪比、复杂随机噪声等,无法获得理想的消噪效果。与其他方法相比,近几年发展起来的时频峰值滤波(TFPF:Time Frequency Peak Filtering)算法具有低信噪比、于非平稳信号等处理能力,但是传统TFPF算法仍存在固定窗长、忽略相邻道间相关性等不足。为此,本文以消减强噪声环境下地震记录中的随机噪声为目的,针对传统TFPF在地震勘探数据处理中的不足,结合径向道变换理论,分别提出了基于平行径向轨线和非线性双曲轨线的时空二维方向轨线TFPF算法,并通过人工合成地震记录以及野外实际地震资料的处理验证了新算法的滤波性能及实用性。本研究基于径向道变换思想,突破传统TFPF处理非线性信号的局限,首次构建平行径向轨线TFPF消噪模型,实现中高频信号的高保真恢复。平行径向轨线TFPF利用地震记录中相邻地震道间反射波的相关性,打破了传统TFPF算法仅沿时间方向滤波的局限性,从根本上减小了由TFPF产生的误差,克服了传统TFPF中采用固定滤波窗长造成不同频率成分反射波的幅值衰减,甚至畸变等缺点,有效实现了中高频反射波的高保真恢复。文中深入探讨了径向轨线方程建立、最优滤波轨线选取、重采样点坐标近似以及样本点插值等关键环节,结合模拟地震记录讨论了不同斜率的径向轨线对消噪结果的影响以及滤波窗长的选取。在平行径向轨线TFPF算法中,沿轨线重采样后随机噪声无明显变化,而有效地震波线性度显著提高,主频明显降低,进而有效降低TFPF估计偏差,使恢复出的地震波幅值与能量得到很好的保持。不同背景噪声下合成地震记录的仿真实验及其与传统TFPF的对比结果表明,在相同窗长条件下,平行径向轨线TFPF方法有效消减随机噪声的同时,恢复出的反射波幅值和能量更接近理想值,有效频率成分(尤其是高频反射同相轴)保持得更完整。同时,该算法对记录中的低频噪声也有较好的消减能力。实际共炮点地震资料的滤波结果表明,平行径向轨线TFPF处理后反射同相轴更清晰,更连续;原本淹没于强随机噪声中的弱反射同相轴的能量及连续性均得到增强而清晰显现。为获得高品质的地震资料及进一步提高信噪比,针对径向轨线与反射同相轴不能完全匹配的局限性,本文充分利用了地震子波的时空相关性并结合反射同相轴在地震记录中的分布形态,从反射波时距关系曲线出发,建立了非线性双曲轨线TFPF去噪模型,有效避免了轨线与同相轴不匹配造成部分能量的衰减,完善了平行径向轨线TFPF消噪模型。本文对模型中的双曲轨线方程建立、最优滤波轨线选取及数据样本点采样等进行了研究。双曲轨线与同相轴的高匹配度最大限度地提高了采样后有效波的线性度,其频率显著降低,滤波窗长更长,消减噪声的能力更强,滤波效果不再受到滤波窗长的严格限制,窗长的选择范围更大。双曲轨线TFPF较径向轨线TFPF的窗长选取更灵活。合成地震记录的消噪结果表明,双曲轨线TFPF算法在低信噪比环境中具有良好的滤波性能。与平行径向轨线TFPF相比,在相同窗长条件下具有更好的噪声消减效果,其恢复的反射波幅值和频带与理想值最为接近,子波能量得到更好的保持,滤波后记录的信噪比大幅提高。对实际共炮点地震资料的处理结果表明,双曲轨线TFPF算法在地震勘探随机噪声压制中更具优越性,恢复出的反射同相轴也更清晰、连贯,其轮廓更平滑,滤波后地震记录的信噪比和分辨率明显提高。两种时空轨线TFPF模型均充分利用了地震波时空相关性,将滤波方向拓展为与反射同相轴形态接近的最优滤波轨线方向,实现有效反射波线性度的优化。滤波轨线的形态与样式直接影响重采样后信号的线性度及主频变化,进而影响TFPF估计偏差与去噪效果,因此最优滤波轨线选取是整个时空轨线TFPF消噪模型中的关键环节。本文针对不同轨线样式分别提出了两种最优滤波轨线选取方法。在平行径向轨线模型中,首次提出通过寻找同相轴上距轴两端点连线最远点来确定最佳滤波轨线。根据几何中点到直线的距离公式分别计算出同相轴上不同点到两端点连线的距离值,连接最大距离点与某一轴端点来获得最优滤波轨线。在双曲轨线模型中,将地震记录中同相轴看成图像边缘,并基于Canny算子边缘检测法实现同相轴位置及走势检测,根据反射波与噪声沿同相轴方向的相关性差异,采用加权均值方法确定轨线曲率变化范围,选取该范围内沿轨线叠加能量最大值所对应的轨线作为最优滤波轨线。
[Abstract]:Seismic exploration, as one of the main exploration methods in geophysical exploration, plays an important role in resource exploration and geological structure research. The seismic data collected in the field often contain a lot of noise, which seriously affects the follow-up inversion and interpretation work. The signal-to-noise ratio (SNR) and the resolution of the data are of great significance to the study of geological structure and the exploration of oil and gas resources.Most of the existing denoising methods at home and abroad are limited by certain assumptions or conditions.Under certain conditions, such as low SNR, complex random noise, etc., ideal denoising effect can not be obtained. TFPF (Time Frequency Peak Filtering) algorithm developed in 1998 has low signal-to-noise ratio (SNR) and non-stationary signal processing ability, but the traditional TFPF algorithm still has some shortcomings, such as fixed window length, ignoring the correlation between adjacent channels and so on. The shortcomings of traditional TFPF in seismic exploration data processing are discussed. Combining with the theory of radial trace transform, the TFPF algorithm of space-time two-dimensional directional trajectory based on parallel radial trajectory and nonlinear hyperbolic trajectory is proposed respectively. The filtering performance and practicability of the new algorithm are verified by processing synthetic seismograms and field seismic data. Based on the idea of radial track transform, this paper breaks through the limitation of traditional TFPF in processing nonlinear signals, and constructs a parallel radial track TFPF denoising model for the first time to realize high fidelity recovery of medium and high frequency signals. The limitation of filtering can reduce the error caused by TFPF fundamentally, overcome the disadvantage of amplitude attenuation or even distortion of reflected waves with different frequency components caused by fixed filtering window length in traditional TFPF, and effectively realize high fidelity recovery of medium and high frequency reflections. Taking, resampling point coordinate approximation and sample point interpolation as key links, the influence of radial trajectories with different slopes on noise reduction and the selection of filtering window length are discussed in combination with simulated seismic records. The simulation results of synthetic seismograms under different background noises and their comparison with traditional TFPF show that the parallel radial trajectory TFPF method can effectively reduce the random noise under the same window length. The amplitude and energy of the reflected wave recovered are closer to the ideal value, and the effective frequency components (especially the high frequency reflection in-phase axis) are more complete. In order to obtain high-quality seismic data and further improve the signal-to-noise ratio, in view of the limitation of the incomplete matching between the radial trajectory and the reflection coaxial, this paper makes full use of the spatial-temporal correlation of seismic wavelet. The TFPF denoising model of nonlinear hyperbolic trajectory is established based on the time-distance relation curve of reflection wave. The TFPF denoising model of parallel radial trajectory is improved by avoiding the partial energy attenuation caused by the mismatch between the trajectory and the phase axis. Establishment, selection of optimal filtering trajectory and sampling of data sample points are studied. The high matching degree between hyperbolic trajectory and phase axis maximizes the linearity of the effective wave after sampling. Its frequency is significantly reduced, the filtering window is longer, the ability of noise reduction is stronger, the filtering effect is no longer strictly limited by the filtering window length and the window length. The results of synthetic seismic data denoising show that the hyperbolic track TFPF algorithm has good filtering performance in low signal-to-noise ratio environment. Compared with the parallel radial track TFPF, the hyperbolic track TFPF has better noise reduction effect under the same window length, and its recovery is opposite. The amplitude and frequency band are the closest to the ideal value, the wavelet energy is better maintained, and the signal-to-noise ratio is greatly improved after filtering. The processing results of real common shot seismic data show that the hyperbolic trajectory TFPF algorithm has more advantages in suppressing random noise in seismic exploration, and the reconstructed reflection coaxiality is clearer and more coherent. Both TFPF models make full use of the spatial-temporal correlation of seismic waves and extend the filtering direction to the optimal filtering trajectory direction close to the reflection in-phase shape, so as to optimize the linearity of the effective reflection wave. In this paper, two optimal filtering trajectory selection methods are proposed for different trajectory patterns. In the parallel radial trajectory model, the optimal filtering trajectory selection is the key link of the whole time-space trajectory TFPF denoising model. For the first time, the optimal filtering trajectory is determined by finding the farthest point of the line connecting the two ends of the coaxial axis. According to the distance formula from the geometric midpoint to the straight line, the distances from different points on the coaxial axis to the two ends of the line are calculated respectively, and the maximum distances are connected with the end points of a certain axis to obtain the optimal filtering trajectory. In seismic records, the in-phase axis is regarded as the edge of the image, and the position and trend of the in-phase axis are detected based on the Canny operator edge detection method. According to the correlation difference between the reflected wave and noise along the in-phase axis, the curvature range of the track is determined by the weighted mean method, and the track corresponding to the maximum superimposed energy along the track is selected. As the optimal filtering trajectory.
【学位授予单位】：吉林大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：P631.4

【相似文献】