基于加权K均值聚类的多属性初至拾取方法

为提高初至拾取方法的准确性和自适应能力，将变异系数加权K均值聚类算法引入初至拾取中。首先提取均方根振幅、相邻道相关性、线积分、振幅谱主频等多种地震属性；然后针对地震属性进行加权K均值聚类，自动识别初至所在时窗；最后结合相位校正法，实现时窗内初至波起跳时间的拾取。在此基础上通过实际数据测试，并与长短时窗能量比法、反向传播神经网络方法对比，验证了本文方法的有效性与可行性。结果表明，基于加权K均值聚类的多属性初至拾取方法能较快速、准确地拾取低信噪比数据的初至，并且无需人为判断时窗，从而提高了拾取的自适应能力。 

First arrival picking method by seismic multi-attribute based on weighted K-means clustering algorithm

For the purpose of improving the accuracy and automation of first arrival picking method, the weighted K-means clustering algorithm is introduced. Firstly, various seismicattributes such as root-mean-squares amplitude, adjacent trace correlation, line integral and dominant frequency of amplitude spectrum are extracted. Then, weighted K-means clustering is performed for seismic attributes to identify the first arrival time window automatically.Finally, combined with phase correction method, this method is applied to realize the pickup of the first arrival time in the time window. The validity and feasibility of the proposed method, which is compared with STA/LTA and BP neural network, are verified by theoretical and practical data tests. The results suggest that the multi-attribute first arrival picking method based on weighted K-means clustering can pick up the first arrival of seismic data with low signal-to-noise ratios quickly and accurately, and enhance the automation of arrival time picking without the artificial identification of time window.