地震数据去噪是地震资料处理的核心环节,去噪的质量直接影响后续处理的精度.实际地震数据中含有多种随机噪声,常见的随机噪声有高斯噪声和椒盐噪声,通常需要采用不同的去噪算法分别进行压制,不可避免会造成有效信号的损失.双边滤波算法能够压制高斯噪声,但是对椒盐噪声不敏感,多级中值滤波算法能够压制椒盐噪声,无法去除高斯噪声.文章将多级中值滤波算法代入双边滤波算法核函数,提出一种基于多级中值的双边滤波算法,能够同时去除地震数据中的高斯噪声和椒盐噪声,更好地保留有效信号.

Deblending of simultaneous-source seismic data is becoming more popular in seismic exploration since it can greatly improve the efficiency of seismic acquisition and reduce acquisition cost. At present, the deblending methods of simultaneous-source seismic data are mainly divided into two types: filtering method and sparse inversion method. Compared with the filtering method, the sparse inversion method has higher precision, but the selection of its parameter value mainly depends on experience, which is not suitable for large-scale seismic data processing. In this paper, an adaptive iterative deblending method based on sparse inversion is proposed. By improving the original iterative solution method of regularization inversion model, the effective signal and blending noise are weakened simultaneously in the iterative process, so that the energy intensity of blending noise is consistent with that of the effective signal in each iterative, so as to ensure the consistency of the regular parameter calculation method of each iteration. By analyzing the distribution of coefficients in the curvelet domain of pseudo-deblending data and blending noise, it is concluded that the value of regular parameters is the maximum amplitude of residual pseudo-deblending data in the curvelet domain multiplied by a coefficient between 0 and 1. In the process of iterative deblending, the regularized parameters are obtained adaptively from the data itself. It not only ensures the accuracy of the calculation results, but also improves the calculation efficiency, which is suitable for large-scale seismic data processing.

The simultaneous source acquisition technology can greatly improve the sampling efficiency. However, compared with the traditional acquisition technology, it may also lead to blended noise which reduces the imaging accuracy. For the 2-D blended data, we usually suppress such noise based on its incoherence in the non-common shot domain. While 3-D blended data contains more information, it has more strongly blended noise and makes it more difficult to construct the blending source operator. To solve these two problems, this paper proposes to separate the 3-D blended data in the Radon domain with sparse constrained inversion which can get higher precision of separation results. Using the GPS time excited by the source to blending and pseudo-deblending the results of the last iterative separation at the common receiver point gather by a long record can process the blending data one receiver by one receiver iteratively rather than the whole data. Such a method does not need to the construct the deblending operator. Tests on synthetic and measured data have proved the feasibility of this method.

Compared with the traditional seismic data acquisition, the mixed source acquisition technology has the advantages of improving the image quality and acquisition efficiency. Reducing the random delay range between single sources in the mixed sources can effectively enhance the efficiency of the acquisition, but it also has a negative impact on the separation of the mixed data. The blending noise in the non-common-source domain data is significantly more concentrated after pseudo-separation, which is difficult to suppress when the random delay time range is small. In this paper, we propose a method to separate the simultaneous source seismic data based on the pulse detection method, and compare it with the iterative multi-level median filtering method. When the time delay range is large, the two methods can both get a good separation result. When the time delay range is small, the method presented in this paper can be more effective, and retain more details of information. The actual data calculation shows that this method can also suppress other random noise to a certain extent.

Full waveform inversion (FWI) reconstructs the underground velocity structures by minimizing the data residual between calculated wavefields and observed wavefileds. The conventional FWI usually uses some local optimization algorithms which lead to a strong dependency on initial velocity model. The objective function corresponding to low-frequency data components has less local minima. Reconstructing low-frequency information from recorded seismic data and using it in FWI can reduce cycle-skipping and thus weaken the dependency of inversion process on initial model. In this paper, based on the conventional frequency down-shifting method, we propose a sparse blind deconvolution-convolution low-frequency data reconstruction method, which can simultaneously update the wavelet and reconstruct the low-frequency components. First, we extract the subsurface reflection impulse responses (SRIR) by solving a Ll norm sparse constraint problem using the Fast Iterative Shrinkage-Thresholding Algorithm (FISTA). Then we test the accuracy of our algorithm, and discuss the effect of wavelet error and noises on the reconstruction result.When the wavelet is inaccurate, we update the amplitude of wavelet by alternately inverting Ll norm constraint and Tikhonov regularization problems, and correct the time-shift error by cross-correlating the direct waves. After that we can get the accurate wavelet and SRIR simultaneously. Then using the reconstructed data successively as observed data, combining it with dynamic random sources and layer-stripping methods, we propose a new strategy for the fast multiscale FWI. We test our method by numerical examples in several cases including blended acquisition cases. The results show that it has good anti-noise property and it can reconstruct valid low frequency components when the observed data lacking low-frequency information. The example using inaccurate wavelet shows that the blind deconvolution-convolution algorithm is able to obtain accurate wavelet and low-frequency data simultaneously. (C) 2017 Elsevier B.V. All rights reserved.

The conventional direct imaging method of microseismic data regards the record as the incident wavefield as well as the scattered wavefield. However, this method cannot highlight the information carried by the wavefield from the deep source. In this paper, assuming that the position and the wavelet of the microseismic event are already known, we propose a reverse time migration (RTM) method to image the structure below the source. The proposed method is similar to the conventional RTM, only the source is underground. During imaging, the deeper wavefield will be more precise than that in the case when regarding the record as the incident wavefield. Hence we can acquire more accurate imaging result. However, the upgoing wavefield from the source underground will bring internal multiple, which could disturb the receiver wavefield. We decompose both the source wavefield as well as the receiver wavefield into the leftgoing and rightgoing parts to suppress this noise. Finally, we use a numerical example to demonstrate the effectiveness of the proposed method. Presentation Date: Wednesday, September 27, 2017 Start Time: 3:05 PM Location: Exhibit Hall C/D Presentation Type: POSTER

The simultaneous source technology has become popular since it can provide a better image much more efficiently. It can effectively improve the efficiency of acquisition by reducing the range of random delay between single source in mixed source, however, it may also bring some negative effects on the deblending of the mixed data. The blending noise contained in non-common-source domain data will be much more concentrated after pseudo-separation, which is difficult to suppress. The trilateral filter, an improvement of the bilateral filter, is based on the ROAD (rank ordered absolute difference). This paper deblends the simultaneous source seismic data based on the trilateral filter. Compared with the multilevel median filter, we can achieve good results by using both two filter when the random delay time range is large, however, when the random delay time range is small, the method presented in this paper can be more effective, and besides, it can also retain more detail information. Presentation Date: Tuesday, September 26, 2017 Start Time: 3:30 PM Location: Exhibit Hall C, E-P Station 2 Presentation Type: EPOSTER

Wave-equation Traveltime Inversion (WTI) is a good method for building background velocity models, and it can provide a good initial velocity model for full waveform inversion (FWI). But, sometimes, the WTI result is not good enough for conventional FWI, while in order to calculate the traveltime difference between synthetic data and recorded data, we have to extract the first arrival waveform and use cross-correlation method. In this paper, we propose to use Single Frequency (SF) waveform to conduct Wave-equation Traveltime Inversion, and combine with frequency multi-scale strategy, which can build a high-precision initial model for conventional FWI, we call it as Single Frequency waveform Wave-equation Traveltime Inversion (SFWTI). The only difference between single frequency waveform of synthetic data and recorded data is traveltime, and it does not need to extract the first arrival waveform. We use the traveltime difference with low frequency waveform to restore the macro structure of velocity models, and then gradually increase the frequency of the seismic data in order to obtain the details of underground structure. Numerical examples show that SFWTI can provide a high-precision initial velocity model for conventional FWI even started with 15Hz. SFWTI+conventional FWI can recover underground detail information and effectively mitigate the cycle skipping problem for FWI. Presentation Date: Wednesday, September 27, 2017 Start Time: 3:05 PM Location: Exhibit Hall C/D Presentation Type: POSTER