基于GPU计算的三维地震断层解

3-D seismic fault interpretation based on GPU computation

1.成都理工大学“油气藏地质及开发工程”国家重点实验室,四川成都610059;2.中国石油天然气集团公司川庆钻探工程有限公司地质勘探研究院,四川成都610051

Zhong Yong,State Key Laboratory of Oil and Gas Reservoir Geology and Exploitation,Chengdu University of Technology,Chengdu 610059,China

应用图像处理方法对地震图像进行预处理后,可直接进行地震断层解释,但基于偏微分方程的图像扩散滤波方法计算效率低,不利于三维地震图像处理。基于GPU(Graphic Process Unit)计算的快速发展,应用GPU加速计算提取三维地震图像的结构张量,由结构张量导向进行地震图像扩散滤波增强预处理;基于三维结构张量识别的三维地震图像不连续区域,采用边缘检测和方差分析方法检测地层和断层区域,再通过设定阈值,提取三维空间中的断层。分别使用CPU和GPU计算处理了样点数为128×128×128和256×256×256的两个三维地震数据体,应用扩散增强滤波技术迭代10次,GPU计算速度分别是CPU计算速度的16.1倍和16.8倍。截取任意三维地震数据体的时间切片，对比了人工拾取断层和计算机自动拾取断层的效果,可以看出,对于较大尺度的断层,人工拾取和计算机自动拾取的结果基本一致,但是对于尺度较小的断层,计算机自动拾取结果更加准确,断层走向更加清晰。

Seismic images can be directly used for seismic fault interpretation after imaging diffusion filtering preprocessing.However,the computation efficiency is low,which is not beneficial for 3-D seismic image.The fast development of Graphic Process Unit (GPU) makes it possible to extract the structure tensor of 3-D seismic image by GPU acceleration computation.Then,by using the structure tensor,diffusion filtering enhancement preprocessing can be carried out on seismic images.Based on the discontinuous area on 3-D seismic image recognized by 3-D structure tensor,edge detection and variance analysis were used to detect the formations and faults.Furthermore,by setting threshold,the faults in 3-D domain were extracted.CPU and GPU computation was applied to process two 3-D seismic data volume with 128×128×128 and 256×256×256 samples.After ten times of iteration with diffusion enhancement filtering,the computation speed of GPU is 16.1 and 16.8 times of CPU computation.The effect of manuallypicking and automaticallypicking was compared by extracting arbitrary time slice of 3-D seismic image volume. The results show that for largescale fault,the two results are almost the same,but for smallerscale fault,automaticallypicking is more accurate,and which can displays clearer fault strike.