基于随机森林方法的地震插值方法研究

Seismic interpolation based on a random forest method

在地震数据的采集过程中,不可避免地会出现地震道缺失或者空间采样不足的情况,这样会产生坏道、缺失道等现象,极大的影响了地震资料质量。想要解决该问题就必须进行地震插值。本文借助于机器学习思想,以无缺失道数据为基础构建机器学习样本集,在此基础上利用随机森林回归预测算法学习各道各时间点振幅与其临近道、时窗内的振幅的统计关系,然后根据临近道数据对缺失道进行补全。将本文所提出方法应用到模型数据与实际采集数据中的缺失道补全处理,均取得良好应用效果,证明本文方法的正确性与有效性。

In the course of any seismic data acquisition, one inevitably encounters instances of empty seismic traces or insufficient spatial sampling, which results in bad sectors and can greatly affect seismic data quality. It is therefore often necessary to undertake seismic trace interpolation to solve this problem. In this paper, a machine learning based method is proposed and applied. This approach requires that the statistical relationship between the amplitude of each trace at each time point and the amplitude of the adjacent trace and time window be derived using a random forest regression prediction algorithm, then the empty trace can be populated according to the adjacent trace data. The method proposed in this paper has achieved good results in the derivation of empty trace values when applied to both model data and actual data, thus proving its validity and effectiveness.