用模式识别技术实现井间小层自动对比

PATTERN RECOGNITION FOR AUTOMATIC THIN LAYER CORRELATION IN CROSS-WELL SEISMOLOGY

地矿部石油物探研究所

The Institute ofGeophysical Prosecting for Petroleum, MGMR

本文介绍一种应用模式识别技术实现井间小层自动对比的方法。首先从井资料提取每一小层的若干主要特征;然后应用模式识别的链匹配算法求得总代价为最小、用概率标记松驰算法求得概率为最大的最优井问地层对比关系,最后,实现井问小层自动对比。野外实际资料的试算结果表明,其效果良好。

The paper suggests to use pattern recognition technique for automatic thin layer correlation in cross-well seismology. In its implementation, Several main features related to each thin layer arc first extracted from well-log data. By means of either pattern recognition-based string-to-string matching procedure, which renders the total cost minimum, or the probability labeling relaxation algorithm which maximizes the probability, the optimum correlation between homologous layers in cross-well seismology is then obtained. Each thin layer down the two wells in a cross-well surrey is correlated as a final procedure. The processing on real data shows that the results are satisfactory.