基于ANN的断层多边形平滑方法研究与实现
|
摘要
随着油田勘探开发的不断深入，对地震资料解释精度的要求越来越高，对解释项目的完成周期要求越来越短，所以地震资料解释技术的发展方向是在高精度的前提下实现自动化的地震资料解释，因此在提取断层以后，对断层进行平滑处理越来越重要。
目前在断层多边形平滑领域已经提出了许多方法，但是至今提出的相关理论和方法尚存在不足之处，在某些情况下仍然无法很好地对断层进行平滑，难以找到一种普遍适应性的断层多边形平滑方法。人工神经网络是计算机科学、信息科学和物理学交互发展形成的一门边缘学科，数字图像处理是人工神经网络目前较新的和最重要的应用领域之一，而图形平滑是图像处理的一个应用。因此，本文设计了一种人工神经网络和断层多边形平滑技术相结合的方法来实现高效率、高精度的断层多边形自动平滑。
本文首先研究了图像平滑及曲线平滑的一般方法、传统多边形平滑方法，比较了各种方法的优缺点。其次，对人工神经网络常用模型进行分析，把模型应用于断层多边形平滑，并取得了明显的效果。最后，针对实际项目的要求设计了基于人工神经网络的断层多边形自动平滑方法，并将其应用到断层多边形平滑及解释系统中，该系统可以完成对有复杂断层分布和噪声干扰的区块，自动提取出的断层多边形进行自动平滑处理，再将人工交互处理结果输出到文件并进行保存等功能，该平滑方法平滑完成的断层多边形光顺、连续以及具有丰富的边缘细节。现今，已完成断层多边形平滑及解释系统的设计并将其应用到实际项目中，且取得了良好的使用效果，提高了断层解释的效率和精度，并缩短了解释周期。
With the deepening development of the oil field exploration, it demands higher andhigher interpretation accuracy and shorter interpretation period for interpreting, so thedevelopment direction of seismic data interpretation technology is to realize automaticinterpretation on condition of high precision, therefore after extracting, the fault smoothing ismore and more important.
At present many methods have been proposed in the fault polygon smooth field, but sofar there are still deficiencies for the related theory and method proposed, and in some casesthe effect is not very well, so it is hard to find a commonly adaptable fault polygon smoothingmethod. Artificial neural network is a borderline subject combined with computer science,information science and interaction of the formation of the development of physics. Thedigital image processing is a new and the most important one of the application in field ofartificial neural network at present.While graphics smooth is an application of the imageprocessing.Therefore, this paper designed a method which combined with artificial neuralnetwork and fault polygon smoothing technique to realize the high efficiency and highprecision after automatically smoothing.
The paper first studied the general smoothing method for the image and curve and thetraditional polygon smoothing method, then compared the advantages and disadvantages ofeach method. Second, the artificial neural network model which is commonly used wasanalyzed, and the model is applied to fault polygon smooth, and have made an obvious effect.Lastly, according to the actual needs of the project it designed fault polygon automaticallysmoothing method based on artificial neural network, and the method is applied to faultpolygon smoothing and interpretation system.Using the system, it can smooth automaticallyfor blocks with complex fault distribution and noise, and extract fault polygon automatically,then save the output result.The practice proved that the result is smooth, continuous and is ofrich edge details.Today, it has completed the design of the fault polygon smooth andexplanation system.This system has been applied to the practical projects and achieved gooduse of effects, and using the system, the precision and efficiency of fault interpretation isimproved and the explanation cycle is shortened.
目前在断层多边形平滑领域已经提出了许多方法，但是至今提出的相关理论和方法尚存在不足之处，在某些情况下仍然无法很好地对断层进行平滑，难以找到一种普遍适应性的断层多边形平滑方法。人工神经网络是计算机科学、信息科学和物理学交互发展形成的一门边缘学科，数字图像处理是人工神经网络目前较新的和最重要的应用领域之一，而图形平滑是图像处理的一个应用。因此，本文设计了一种人工神经网络和断层多边形平滑技术相结合的方法来实现高效率、高精度的断层多边形自动平滑。
本文首先研究了图像平滑及曲线平滑的一般方法、传统多边形平滑方法，比较了各种方法的优缺点。其次，对人工神经网络常用模型进行分析，把模型应用于断层多边形平滑，并取得了明显的效果。最后，针对实际项目的要求设计了基于人工神经网络的断层多边形自动平滑方法，并将其应用到断层多边形平滑及解释系统中，该系统可以完成对有复杂断层分布和噪声干扰的区块，自动提取出的断层多边形进行自动平滑处理，再将人工交互处理结果输出到文件并进行保存等功能，该平滑方法平滑完成的断层多边形光顺、连续以及具有丰富的边缘细节。现今，已完成断层多边形平滑及解释系统的设计并将其应用到实际项目中，且取得了良好的使用效果，提高了断层解释的效率和精度，并缩短了解释周期。
With the deepening development of the oil field exploration, it demands higher andhigher interpretation accuracy and shorter interpretation period for interpreting, so thedevelopment direction of seismic data interpretation technology is to realize automaticinterpretation on condition of high precision, therefore after extracting, the fault smoothing ismore and more important.
At present many methods have been proposed in the fault polygon smooth field, but sofar there are still deficiencies for the related theory and method proposed, and in some casesthe effect is not very well, so it is hard to find a commonly adaptable fault polygon smoothingmethod. Artificial neural network is a borderline subject combined with computer science,information science and interaction of the formation of the development of physics. Thedigital image processing is a new and the most important one of the application in field ofartificial neural network at present.While graphics smooth is an application of the imageprocessing.Therefore, this paper designed a method which combined with artificial neuralnetwork and fault polygon smoothing technique to realize the high efficiency and highprecision after automatically smoothing.
The paper first studied the general smoothing method for the image and curve and thetraditional polygon smoothing method, then compared the advantages and disadvantages ofeach method. Second, the artificial neural network model which is commonly used wasanalyzed, and the model is applied to fault polygon smooth, and have made an obvious effect.Lastly, according to the actual needs of the project it designed fault polygon automaticallysmoothing method based on artificial neural network, and the method is applied to faultpolygon smoothing and interpretation system.Using the system, it can smooth automaticallyfor blocks with complex fault distribution and noise, and extract fault polygon automatically,then save the output result.The practice proved that the result is smooth, continuous and is ofrich edge details.Today, it has completed the design of the fault polygon smooth andexplanation system.This system has been applied to the practical projects and achieved gooduse of effects, and using the system, the precision and efficiency of fault interpretation isimproved and the explanation cycle is shortened.
引文
[1]平丽．图像平滑处理方法的比较研究[J]．信息技术．2010(1):65~68．
[2] Rugumira Georgia,GAO Guaijing,WANG Lingling.Based on the direction of the templateto maintain smooth edges seismic images of the discontinuity detection[J].Progress ingeophysics,2008(5):1398-1405.
[3]张义宽．计算机图形学[M]．西安电子科技大学出版社，2004：93~100．
[4]滕召荣，蒋天发．邻域平均法对矢量图平滑处理[J]．现代电子技术．2009(14)：75~77．
[5]王双．一种改进的中值滤波算法[J]．电子技术．2007：120~123．
[6]林芳．Bézier曲线的光滑拼接[J]．渭南师范学院学报．2011(5)：3~5．
[7]赵罡，王爱增，穆国旺．基于小波的非均匀B样条曲线自动光顺算法[J]．计算机辅助设计与图形学．2010，22(11)：1925~1931．
[8] Mark J.1.Orr.Introduetion to Ralial Basis Funetion Networks.Center for Cognitive Seienee,University of Edinburgh,Buccleueh Plaee,Edinburgh EH89LW,Scotland,1996:54-60.
[9]隋英，付春菊，高兴燕．人工神经网络研究的发展与应用[J]．大众科技，2010(5)：20，34．
[10]李凯．对人工神经网络的冷思考[J]．东方企业文化·公司与产业，2010(3)：124．
[11]王爽，张鹰，吕瑞霞．BP神经网络的算法改进及应用[J]．人工智能及识别技术．2009，5(4)：933~935．
[12]周政．BP神经网络的发展现状综述[J]．山西电子技术．2008(2)：90~92．
[13]凌卫新，郑启伦，陈琼．模块化神经网络的系统结构和学习算法[J]，计算机工程与应用．2001，(19)：13~15．
[14]李安新．BP神经网络研究与硬件实现[D]．山东科技大学．2010．
[15]张琛．BP神经网络模型优化研究[J]．吉林省教育学院学报．2011，7(27)：14~152．
[16] Rumethart D E.Learning Representation by BP Errors[J].Nature,1986,7:149-154.
[17] Bertalan Bodri.A neural-network model for earthquake occurrence[J].Journal ofGeodynamics32,2001:289-310.
[18]高宁，张建中．MATLAB在RBF神经网络模型中的应用[J]．农业网络信息．2009(2)：110~113．
[19] Mooky and Darken,C..Fast learning in networks of loeallytuned proeessing units.NeuralComputation,1989,1:281-294.
[20]肖健华．智能模式识别方法[M]．华南理工大学出版社，2006：61~94．
[21]岳彩青，常青美．基于聚类分析的RBF网络建模方法及应用的研究[J]．计算机仿真，2006，22(1)：18~25．
[22] Tyler Holeomb and Manfred Morari. Iocal training for radial basis function networks:towards solving the hidden uint Problem Ameriean Control Conference,1993,2331-2335.
[23]田佳．RBF神经网络在油田试井系统中的应用[J]．吉林工程技术师范学院学报．2010(3)：78~80．
[24]张飞，段志峰．RBF网络在边坡稳定性中的应用研究[J]．西部探矿工程．2011(8)：20~25．
[25]吴群，陈晓辉，李东普等．改进RBF神经网络在降水量预测中的应用[J]．信阳师范学院学报．2011(2)：268~271．
[26]刘雪峰，张宏立．基于MATLAB的RBF神经网络在分类中的应用[J]．内江科技．2010(7)：137~137．
[27]王宗彦，梁远蕾，李奇敏等．断层数据三维重构技术的研究进展[J]．北京：工程图学学报，2002(1)：125~130．．
[28] YANG Yong, ZHAO Yuzhen. A denoising smoothing algorithm of image edgepreserving[J]. Journal of Daqing Petroleum Institute,2008,32(5):78-80.
[29]罗斌，汪炳权．三维表面平滑约束断层图象形状插值[J]．北京：中国图象图形学报，1996，1(4)：294~296．
[30]王发牛．基于偏微分方程的图像平滑技术及其应用研究[D]．安徽大学，2002．
[31]刘彦花，叶国华．矢量曲线抽稀算法分析[J]．武汉：城市勘测．2001(4)：1~4．
[32]朱述龙．一种新的曲线光滑算法一方向判别法[J]．解放军测绘学院学报．1993(3)：49~51．
[33]梁栋．基于图像区域特征的边缘检侧方法研究与实现[D]．东北石油大学，2011．
[34]刘丽君，熊泽林，骆婷．一种基于边缘检测的图像平滑算法[J]．中国水运．2009，9(4)：92~93．
[35] Luo Y, Marhoon M, Dossary S A, Alfaraj M A. Edge preserving smoothing andapplications[J]. The leading edge.2002,21:136~158.
[36] BONDAR I.Seismic horizon detection using image processing algorithms[J].GeophysicalProspecting,1992,40:785-800.
[37] He Yu ao,Hu Xianzhong, Zhan Sheng. Predicting Seismic Response of Structures byArtificial Neural Network[J].Transactions of Tianjin University,1996,2(2):36~39
[38] Tang Shi-wei,LI Ming-juan,Fan Ying-ying. Research on Fault Polygon Extraction andSmoothing Method[J].20112nd International Conference on Mechanic Automation andControl Engineering,MACE2011,2011:2474-2476..
[39]张小磊，图像边缘处理技术方法的研究及应用[D]．南京信息工程大学，2009．
[40] SRINIVASAN V, BHATIA P. Edge detection using a neural network[J]. PatternRecognition,1994,27(12):1653-1662.
[41] Li Weiqing,Wang Chengbiao,Wang Qun,et al. An edge detection method based onoptimized BP neural network[J]. International Symposium on Information Science andEngineering,2008(2):40-44.
[42]张德丰．MATLAB神经网络应用设计[M]．机械工程出版社，2009：4~5．
[43]翟亮亮，王连堂，王俊．基于Levenberg-Marquardt算法的油田产量预测模型[J].石油地质与工程．2010(5):63~65.
[44] WANG W B,ZHENG P,LI J Y. Application of BP NN and RBF NN in modelingactivated sludge system[J].Transactions of Tianjin University,2003,9(3):235-240.
[45]周开利，康耀红．神经网络模型及其MATLAB仿真程序设计[M]．2005．
[46]黄先锋，程晓光，张帆等．基于边长比约束的离散点准确边界追踪算法[J]．信息科学．2009，34(6)：688~691．
[47] Chitwong S,Cheevasuit F,Dejhan K,Mitatha S,Nokyoo C,Paungma T.Segmentation onedge preserving smoothing image based on graph theory[C].Geoscience and RemoteSens-ing Symposium,2000.Proceedings.IGARSS2000.IEEE2000International.2000,2:621-623.
[48]卢洋．分数阶微分应用于图像边缘检测的研究[D]．大连交通大学，2009．
[49]王兆湖，刘财，王建民等．基于层位解释的断层多边形自动组合技术[J]．石油地球物理勘探，2009，44(1)：104~106．
[50] Tang Shi-wei, Fan Ying-ying,Li Ming-juan. Design on a method for interactive editingfault polygon[J]. Advanced In Control Engineering and Information Science(CEIS2011),2011:3584-3589.
[51]冯永强．图像边缘信息处理技术研究及应用[D]．大庆石油学院，2007．
[2] Rugumira Georgia,GAO Guaijing,WANG Lingling.Based on the direction of the templateto maintain smooth edges seismic images of the discontinuity detection[J].Progress ingeophysics,2008(5):1398-1405.
[3]张义宽．计算机图形学[M]．西安电子科技大学出版社，2004：93~100．
[4]滕召荣，蒋天发．邻域平均法对矢量图平滑处理[J]．现代电子技术．2009(14)：75~77．
[5]王双．一种改进的中值滤波算法[J]．电子技术．2007：120~123．
[6]林芳．Bézier曲线的光滑拼接[J]．渭南师范学院学报．2011(5)：3~5．
[7]赵罡，王爱增，穆国旺．基于小波的非均匀B样条曲线自动光顺算法[J]．计算机辅助设计与图形学．2010，22(11)：1925~1931．
[8] Mark J.1.Orr.Introduetion to Ralial Basis Funetion Networks.Center for Cognitive Seienee,University of Edinburgh,Buccleueh Plaee,Edinburgh EH89LW,Scotland,1996:54-60.
[9]隋英，付春菊，高兴燕．人工神经网络研究的发展与应用[J]．大众科技，2010(5)：20，34．
[10]李凯．对人工神经网络的冷思考[J]．东方企业文化·公司与产业，2010(3)：124．
[11]王爽，张鹰，吕瑞霞．BP神经网络的算法改进及应用[J]．人工智能及识别技术．2009，5(4)：933~935．
[12]周政．BP神经网络的发展现状综述[J]．山西电子技术．2008(2)：90~92．
[13]凌卫新，郑启伦，陈琼．模块化神经网络的系统结构和学习算法[J]，计算机工程与应用．2001，(19)：13~15．
[14]李安新．BP神经网络研究与硬件实现[D]．山东科技大学．2010．
[15]张琛．BP神经网络模型优化研究[J]．吉林省教育学院学报．2011，7(27)：14~152．
[16] Rumethart D E.Learning Representation by BP Errors[J].Nature,1986,7:149-154.
[17] Bertalan Bodri.A neural-network model for earthquake occurrence[J].Journal ofGeodynamics32,2001:289-310.
[18]高宁，张建中．MATLAB在RBF神经网络模型中的应用[J]．农业网络信息．2009(2)：110~113．
[19] Mooky and Darken,C..Fast learning in networks of loeallytuned proeessing units.NeuralComputation,1989,1:281-294.
[20]肖健华．智能模式识别方法[M]．华南理工大学出版社，2006：61~94．
[21]岳彩青，常青美．基于聚类分析的RBF网络建模方法及应用的研究[J]．计算机仿真，2006，22(1)：18~25．
[22] Tyler Holeomb and Manfred Morari. Iocal training for radial basis function networks:towards solving the hidden uint Problem Ameriean Control Conference,1993,2331-2335.
[23]田佳．RBF神经网络在油田试井系统中的应用[J]．吉林工程技术师范学院学报．2010(3)：78~80．
[24]张飞，段志峰．RBF网络在边坡稳定性中的应用研究[J]．西部探矿工程．2011(8)：20~25．
[25]吴群，陈晓辉，李东普等．改进RBF神经网络在降水量预测中的应用[J]．信阳师范学院学报．2011(2)：268~271．
[26]刘雪峰，张宏立．基于MATLAB的RBF神经网络在分类中的应用[J]．内江科技．2010(7)：137~137．
[27]王宗彦，梁远蕾，李奇敏等．断层数据三维重构技术的研究进展[J]．北京：工程图学学报，2002(1)：125~130．．
[28] YANG Yong, ZHAO Yuzhen. A denoising smoothing algorithm of image edgepreserving[J]. Journal of Daqing Petroleum Institute,2008,32(5):78-80.
[29]罗斌，汪炳权．三维表面平滑约束断层图象形状插值[J]．北京：中国图象图形学报，1996，1(4)：294~296．
[30]王发牛．基于偏微分方程的图像平滑技术及其应用研究[D]．安徽大学，2002．
[31]刘彦花，叶国华．矢量曲线抽稀算法分析[J]．武汉：城市勘测．2001(4)：1~4．
[32]朱述龙．一种新的曲线光滑算法一方向判别法[J]．解放军测绘学院学报．1993(3)：49~51．
[33]梁栋．基于图像区域特征的边缘检侧方法研究与实现[D]．东北石油大学，2011．
[34]刘丽君，熊泽林，骆婷．一种基于边缘检测的图像平滑算法[J]．中国水运．2009，9(4)：92~93．
[35] Luo Y, Marhoon M, Dossary S A, Alfaraj M A. Edge preserving smoothing andapplications[J]. The leading edge.2002,21:136~158.
[36] BONDAR I.Seismic horizon detection using image processing algorithms[J].GeophysicalProspecting,1992,40:785-800.
[37] He Yu ao,Hu Xianzhong, Zhan Sheng. Predicting Seismic Response of Structures byArtificial Neural Network[J].Transactions of Tianjin University,1996,2(2):36~39
[38] Tang Shi-wei,LI Ming-juan,Fan Ying-ying. Research on Fault Polygon Extraction andSmoothing Method[J].20112nd International Conference on Mechanic Automation andControl Engineering,MACE2011,2011:2474-2476..
[39]张小磊，图像边缘处理技术方法的研究及应用[D]．南京信息工程大学，2009．
[40] SRINIVASAN V, BHATIA P. Edge detection using a neural network[J]. PatternRecognition,1994,27(12):1653-1662.
[41] Li Weiqing,Wang Chengbiao,Wang Qun,et al. An edge detection method based onoptimized BP neural network[J]. International Symposium on Information Science andEngineering,2008(2):40-44.
[42]张德丰．MATLAB神经网络应用设计[M]．机械工程出版社，2009：4~5．
[43]翟亮亮，王连堂，王俊．基于Levenberg-Marquardt算法的油田产量预测模型[J].石油地质与工程．2010(5):63~65.
[44] WANG W B,ZHENG P,LI J Y. Application of BP NN and RBF NN in modelingactivated sludge system[J].Transactions of Tianjin University,2003,9(3):235-240.
[45]周开利，康耀红．神经网络模型及其MATLAB仿真程序设计[M]．2005．
[46]黄先锋，程晓光，张帆等．基于边长比约束的离散点准确边界追踪算法[J]．信息科学．2009，34(6)：688~691．
[47] Chitwong S,Cheevasuit F,Dejhan K,Mitatha S,Nokyoo C,Paungma T.Segmentation onedge preserving smoothing image based on graph theory[C].Geoscience and RemoteSens-ing Symposium,2000.Proceedings.IGARSS2000.IEEE2000International.2000,2:621-623.
[48]卢洋．分数阶微分应用于图像边缘检测的研究[D]．大连交通大学，2009．
[49]王兆湖，刘财，王建民等．基于层位解释的断层多边形自动组合技术[J]．石油地球物理勘探，2009，44(1)：104~106．
[50] Tang Shi-wei, Fan Ying-ying,Li Ming-juan. Design on a method for interactive editingfault polygon[J]. Advanced In Control Engineering and Information Science(CEIS2011),2011:3584-3589.
[51]冯永强．图像边缘信息处理技术研究及应用[D]．大庆石油学院，2007．
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |