劉劍 秦飛龍 成亞麗
摘要:采集的野外地震數(shù)據(jù)伴隨有隨機(jī)噪聲干擾,需要將其消除。軟硬閾值法能夠壓制地震數(shù)據(jù)的噪聲信號(hào),但是降噪效果并不理想。因此,提出了一種改進(jìn)的軟硬閾值算法用于地震數(shù)據(jù)降噪。首先利用軟硬閾值法原理構(gòu)建了一種新的閾值降噪法,并對(duì)新算法相關(guān)特性進(jìn)行了研究,通過(guò)仿真實(shí)驗(yàn)確定了新閾值算法的小波基為sym 3,利用均方差和信噪比對(duì)新閾值降噪法的降噪效果進(jìn)行了評(píng)價(jià)。最后,將新閾值降噪法用于實(shí)際地震數(shù)據(jù)降噪,結(jié)果發(fā)現(xiàn)新閾值降噪法能夠去除地震數(shù)據(jù)中的隨機(jī)噪聲,降噪效果較軟硬閾值法更理想。
關(guān)鍵詞:地震數(shù)據(jù);小波變換;軟硬閾值法;新閾值法;降噪
中圖分類號(hào):P631.443
文獻(xiàn)標(biāo)志碼:A文章編號(hào):1000-582X(2022)02-052-06
Abstract: The collected field seismic data are accompanied by random noise interference, which needs to be eliminated by wavelet threshold denoising method. However, the soft and hard thresholding methods have some shortcomings in noise reduction. In this paper, a new threshold denoising algorithm is proposed to perform seismic data denoising. Firstly, a new threshold denoising method is designed based on the principle of soft and hard threshold method, and the relevant characteristics of the new algorithm are studied. Then, the wavelet base of the new threshold algorithm is defined as sym 3, and the denoising effect of the new method is evaluated by root mean square error (RMSE) and signal-to-noise ratio (SNR) through simulation experiments. Finally, the new threshold denoising method is applied to actual seismic data denoising. The results show that the improved threshold denoising method can remove random interference in seismic data, and the denoising effect is better than that of the old soft and hard threshold method.
Keywords: seismic data; wavelet transform; soft and hard threshold method; new threshold method; denoising
地質(zhì)勘查能夠給社會(huì)帶來(lái)重要的能源資源。然而,在野外地質(zhì)勘查中,各種不可預(yù)知的因素給采集到的地震數(shù)據(jù)帶來(lái)隨機(jī)噪聲干擾,使地震剖面分辨率不高,不利于地震數(shù)據(jù)解釋,因而需要對(duì)地震數(shù)據(jù)的隨機(jī)噪聲進(jìn)行消除[1]。為了去除隨機(jī)噪聲,學(xué)者們根據(jù)隨機(jī)噪聲與有效信號(hào)的差異,提出了一些隨機(jī)噪聲消除方法,如K-L變換、傅里葉變換、傾斜疊加法等。這些降噪方法需要數(shù)據(jù)的一些先驗(yàn)條件,但實(shí)際上地震數(shù)據(jù)受到的人為影響、機(jī)器影響、地質(zhì)環(huán)境影響等干擾是無(wú)法預(yù)知的。因此,這些算法降噪效果不佳[2]。近年來(lái),盲源分離[3]中的ICA算法[4]、FASTICA算法[5]、JADE算法[6]等也廣泛應(yīng)用于地震數(shù)據(jù)降噪,但是盲源分離方法要求各個(gè)信號(hào)是相互獨(dú)立而且高斯信號(hào)最多為一個(gè),彼此之間的混合不復(fù)雜[2]。地震數(shù)據(jù)埋藏較深,受到干擾影響嚴(yán)重,各種數(shù)據(jù)信息并非線性混合,利用盲源分離進(jìn)行地震數(shù)據(jù)降噪存在缺陷。在隨機(jī)噪聲降噪中,小波降噪具有一定的處理優(yōu)勢(shì),因?yàn)樾〔ㄗ儞Q不僅具有良好的時(shí)頻局部變換特征,還具有“時(shí)間頻率”窗口自適應(yīng)特征,號(hào)稱信號(hào)降噪的“顯微鏡”[7]。小波變換降噪主要通過(guò)小波分解把地震數(shù)據(jù)分解到低頻和高頻不同尺度空間上,通過(guò)選取合理閾值方法對(duì)各尺度上的小波系數(shù)進(jìn)行噪聲壓制,然后利用小波重構(gòu)變換恢復(fù)原始信號(hào)[8]。在閾值處理中,應(yīng)用最廣泛的是Donoho軟硬閾值降噪算法[9],然而該類閾值算法由于算法特性使降噪具有一定的缺陷性。軟閾值降噪算法通過(guò)恒定的方式壓縮小波系數(shù),會(huì)丟失某些有效的高頻信號(hào);硬閾值降噪算法處理后的小波系數(shù)在閾值處不連續(xù),給重構(gòu)信號(hào)帶來(lái)振蕩,降噪后的信號(hào)不光滑。因此直接利用軟硬閾值降噪效果不理想[10]。但是軟閾值法處理后的小波系數(shù)具有連續(xù)性,硬閾值能夠避免軟閾值法以恒定的方式壓縮小波系數(shù)的影響[10]。因此,筆者在軟硬閾值降噪原理基礎(chǔ)上,結(jié)合其降噪的優(yōu)點(diǎn),設(shè)計(jì)了一種改進(jìn)的軟硬閾值降噪方法,以彌補(bǔ)軟硬閾值降噪的不足,提高地震數(shù)據(jù)降噪效果。
1 新閾值降噪算法
1.1 算法基礎(chǔ)
3 實(shí)際地震數(shù)據(jù)降噪處理
將提出的新閾值函數(shù)用于實(shí)際地震數(shù)據(jù)降噪處理,數(shù)據(jù)來(lái)自中國(guó)地質(zhì)調(diào)查局計(jì)劃項(xiàng)目(1212010916040),數(shù)據(jù)樣本采樣間隔道號(hào)為1,采樣點(diǎn)數(shù)為6 000,采樣間隔為1 ms。為了便于顯示,截取其中的1~111道地震信號(hào)數(shù)據(jù)進(jìn)行降噪處理(圖4(a))。由圖4(a)知,原始地震數(shù)據(jù)剖面被隨機(jī)噪聲干擾嚴(yán)重,分辨率低,看不出地震數(shù)據(jù)形態(tài),不利后期地質(zhì)解釋。因此,選取小波基為sym 3,3層小波分解。利用小波變換中的軟、硬閾值函數(shù)對(duì)原始地震數(shù)據(jù)進(jìn)行降噪,降噪后的信號(hào)分別如圖4(b)(c)所示。由圖4(b)(c)知,軟、硬閾值函數(shù)能夠去除地震數(shù)據(jù)大部分干擾噪聲,地震分辨率有所改善,然而降噪效果不理想,仍然存在大量噪聲影響地震剖面,分辨不高。利用所提出的新閾值函數(shù)對(duì)原始地震數(shù)據(jù)進(jìn)行降噪處理,降噪后的信號(hào)結(jié)果如圖4(d)所示。由圖4(d)知,整個(gè)降噪后的地震剖面數(shù)據(jù)雙曲線特征明顯,紋理清晰,分辨率高,幾乎所有干擾噪聲均被移除,表明新閾值函數(shù)降噪效果比軟、硬閾值函數(shù)降噪效果更理想。
4 結(jié) 論
在軟、硬閾值函數(shù)降噪基礎(chǔ)上提出了一種新閾值降噪函數(shù),結(jié)論如下:
1)新的閾值降噪函數(shù)同時(shí)具備軟、硬閾值降噪函數(shù)功能,其優(yōu)點(diǎn)是避免了軟閾值函數(shù)在恒定偏差上的影響,也避免了硬閾值函數(shù)在閾值處的不連續(xù)性影響;
2)通過(guò)仿真實(shí)驗(yàn)確定了新閾值函數(shù)的小波基為sym 3,信噪比(SNR)和均方根誤差(RMSE)結(jié)果表明新閾值函數(shù)降噪效果更好;
3)在實(shí)際數(shù)據(jù)降噪處理中,新閾值函數(shù)能夠去除地震數(shù)據(jù)的各類干擾噪聲,降噪后的地震剖面分辨率高,對(duì)比軟、硬閾值函數(shù)結(jié)果表明新閾值函數(shù)降噪更有效。
參考文獻(xiàn):
[1] 秦飛龍. 二維盲小波算法及其在金屬地震數(shù)據(jù)降噪處理中的研究[D]. 成都: 成都理工大學(xué), 2013.
Qin F L. 2D blind-wavelet algorithm and its application in the processing of de-noising on metal seismic data[D]. Chengdu: Chengdu University of Technology, 2013. (in Chinese)
[2] Qin F L, Liu J, Yan W Y. The improved ICA algorithm and its application in the seismic data denoising[J]. Journal of Chongqing University English Edition, 2018, 17(4): 162-170.
[3] 刁瑞, 吳國(guó)忱, 尚新民, 等. 地面陣列式微地震數(shù)據(jù)盲源分離去噪方法[J]. 物探與化探, 2017, 41(3): 521-526.
Diao R, Wu G C, Shang X M, et al. The blind separation denoising method for surface array micro-seismic data[J]. Geophysical and Geochemical Exploration, 2017, 41(3): 521-526. (in Chinese)
[4] Safont G, Salazar A, Vergara L, et al. Nonlinear estimators from ICA mixture models[J]. Signal Processing, 2019, 155: 281-286.
[5] Jiang S C, Lin P, Chen Y M, et al. Mixed-signal extraction and recognition of wind turbine blade multiple-area damage based on improved Fast-ICA[J]. Optik, 2019, 179: 1152-1159.
[6] Piotrowski A P, Napiorkowski J J. Step-by-step improvement of JADE and SHADE-based algorithms: success or failure?[J]. Swarm and Evolutionary Computation, 2018, 43: 88-108.
[7] 程魯, 秦飛龍, 張津, 等. 改進(jìn)的Mallat算法在金屬地震數(shù)據(jù)降噪處理中的應(yīng)用[J]. 桂林理工大學(xué)學(xué)報(bào), 2017, 37(4): 602-607.
Cheng L, Qin F L, Zhang J, et al. Application of improved Mallat algorithm in noise reduction processing of metal seismic data[J]. Journal of Guilin University of Technology, 2017, 37(4): 602-607. (in Chinese)
[8] 宋立業(yè), 周樂(lè), 劉昕明. 基于鄰域相關(guān)性多閾值新函數(shù)尋優(yōu)法的小波降噪分析[J]. 信息與控制, 2019, 48(1): 57-64.
Song L Y, Zhou L, Liu X M. Wavelet denoising analysis based on an optimization method of multi-threshold new function with neighborhood correlation[J]. Information and Control, 2019, 48(1): 57-64. (in Chinese)
[9] Donoho D L, Johnstone I M. Ideal spatial adaptation by wavelet shrinkage[J]. Biometrika, 1994, 81(3): 425-455.
[10] 劉書(shū)俊, 李生林, 蔣明, 等. 一種基于加權(quán)平均的改進(jìn)型小波閾值降噪算法[J]. 化工自動(dòng)化及儀表, 2017, 44(3): 239-242, 318.
Liu S J, Li S L, Jiang M, et al. An improved wavelet threshold de-noising algorithm based on weighted average method[J]. Control and Instruments in Chemical Industry, 2017, 44(3): 239-242, 318. (in Chinese)
[11] Wang G X, Chen L, Guo S, et al. Application of a new wavelet threshold method in unconventional oil and gas reservoir seismic data denoising[J]. Mathematical Problems in Engineering, 2015, 2015: 1-7.
[12] Bayer F M, Kozakevicius A J, Cintra R J. An iterative wavelet threshold for signal denoising[J]. Signal Processing, 2019, 162: 10-20.
[13] Cotronei M, Rossini M, Sauer T, et al. Filters for anisotropic wavelet decompositions[J]. Journal of Computational and Applied Mathematics, 2019, 349: 316-330.
[14] 成樞, 馬衛(wèi)驕, 牛英杰, 等. 融合加權(quán)平均的改進(jìn)小波閾值函數(shù)的測(cè)量數(shù)據(jù)處理[J]. 中國(guó)科技論文, 2018, 13(15): 1811-1816.
Cheng S, Ma W J, Niu Y J, et al. Measurement data processing of improved wavelet threshold function with weighted average[J]. China Sciencepaper, 2018, 13(15): 1811-1816. (in Chinese)
[15] 周仲禮, 秦飛龍, 夏歡歡, 等. 盲小波算法在金屬礦床地震資料去噪處理中的應(yīng)用[J]. 成都理工大學(xué)學(xué)報(bào)(自然科學(xué)版), 2013, 40(2): 120-124.
Zhou Z L, Qin F L, Xia H H, et al. Application of blind wavelet algorithm to de-noising of the metallic ore deposit seismic data[J]. Journal of Chengdu University of Technology(Science & Technology Edition), 2013, 40(2): 120-124. (in Chinese)
(編輯 羅 敏)