基于改进蚁群追踪策略的地震层位自动识别方法  

Seismic horizon automatic identification based on ant colony tracking strategy

作　　者：殷文 李援 郭加树[3] 张琳 朱剑兵[4] 李长红[4] 

机构地区：[1]中国石油大学(北京)克拉玛依校区,新疆克拉玛依834000 [2]东营市人力资源和社会保障局信息中心,山东东营257091 [3]中国石油大学(华东)计算机与通信工程学院,山东青岛266580 [4]中国石化胜利油田有限公司物探研究院,山东东营257022

出　　处：《石油地球物理勘探》2017年第3期553-561,共9页Oil Geophysical Prospecting

基　　金：国家重点基础研究发展计划("973"计划)项目(2013CB228604)资助

摘　　要：目前层位拾取和标定更多地是依靠人工或机器辅助的方式进行,如同相轴追踪、基于神经网络和图像边缘提取技术的层位拾取,业界的地震层位追踪方法偏重于三维层位追踪和剖面自动追踪算法。存在效率低、需要人为指定种子点、训练追踪时间长等缺陷,且每次只能完成一个层位(面)追踪,无法对层间关系进行解释。为此,提出了基于改进蚁群追踪策略的地震层位自动识别方法。首先采用改进的加权中值滤波对地震数据进行预处理,在增强同相轴连续性的基础上,减少了传统种子点生长算法中的人为参与;基于蚁群算法的基本思想,引入支持向量机(SVM)技术对地震数据进行分类处理,同时综合考虑地震振幅、瞬时相位、层位倾角、信息素浓度等多种信息,纳入蚁群层位追踪的评价函数,采用改进的蚁群搜索算法实现地震层位自动追踪。通过对追踪区域进行划分,实现区域内小层加密,提供多种合理的层位加密策略,完成地震层位追踪后期处理。实际资料应用结果检验了方法的有效性,为层序地层分析、沉积体系域解释、精细储层描述提供了技术思路。Nowadays horizon picking and identifications are dependent on human being or machine assistance methods,such as event tracking,horizon picking based on neural network and image-edge extraction.For this aspect,researchers in the industry belong to applying 3D horizon tracking and section automatic tracking algorithms.These algorithms have some defects,for example they have low efficiency,need seed points,and spend a long time for training and tracking.Only one horizon can be tracked,so there is no relationship analysis between horizons.Therefore we propose in this paper a horizon automatic recognition based on improved ant colony tracking strategy.First,seismic data is preprocessed by the improved weighted median filter to enhance event continuity rather than seed point growth.Then,based on the basic idea of ant colony algorithm,seismic data is classified and processed on support vector machine(SVM).More information is added into ant colony tracking as evaluation functions,such as seismic amplitude,instantaneous phase,dip angle,and pheromone concentration.Finally the improved ant colony search algorithm achieves horizon automatic tracking.In the post processing,a tracking area is divided into a few segments,small horizons can be encrypted,and this would provide a variety of reasonable horizon encryption strategies.Our application proves the validity of the proposed approach,which may provide great helps for sequence stratigraphy analysis,sedimentary system tract interpretation,and fine reservoir description.

关 键 词：层位自动追踪 地震数据预处理 SVM回归分析 评价函数 改进蚁群算法 信息素更新 

分 类 号：P631[天文地球—地质矿产勘探]