论文部分内容阅读
在岩性地震勘探工作中,常常要对一组相似地震道间出现的微妙特征变化作出评价。比较典型的是在一组地震道的垂向时间轴上各取一个相同起讫时间的时窗,在横向上该道集形成的同相轴基本是水平的。主元分析利用了地震道集之间信息的冗余度,先决定统计特征,然后通过变换减少特征的相关性。以往工作表明,经过这种处理,地震数据量可以减少到原始数据量的10％左右.用求出的主元相关系数可以在有监督或无监督方式下实现地震道的精确分组。如果有一口或多口测井资料可供利用,就可以把井旁地震资料作为类中心,其它道与之比较而完成聚类。
In lithological exploration work, it is often necessary to evaluate the subtle feature changes that occur in a group of similar seismic traces. Typically, a time window of the same origin and ending time is taken on the vertical time axis of a series of seismic traces, and the horizontal axis of the gathers formed in the horizontal direction is substantially horizontal. Principal component analysis takes advantage of the redundancy of information between trace gatherings, decides on statistical features first, and then reduces the correlation of features by transformation. Previous work shows that after this processing, the amount of seismic data can be reduced to about 10% of the original data, and the obtained principal component correlation coefficients can be used to accurately classify the seismic traces under supervised or unsupervised mode. If there is one or more log data available for use, the wellsite seismic data can be used as a cluster center, and other paths can be compared to complete clustering.