论文部分内容阅读
三分量记录可使我们弄清不同波系的偏振参数。利用这些参数,就可得到许多偏振滤波器,尤其是那些应用协方差矩阵的偏振滤波器。虽然如此,但这些滤波器仅考虑了三分量效果,如果在不同的波之间有干扰的话,那么这些滤波器基本上无效。为了克服这种不利因素,提出了一种模式识别的方法。这种方法采用了 Karhunen-Lovève(KL)变换,其主要由以下两步组成:(1)选择作模式识别的波形,波形用 KL 的主特征向量所形成的矢量子空间来表示。这是一个识别步骤。(2)将整个地震道投影到这个子空间上,即进行滤波处理。文中提出可将上述方法应用于三分量偏移距 VSP。列举的例子表明,我们可以从所有道中提取与某个已知的模型波形相似的信号,如:第一个下行 P 波。
Three-component recording allows us to clarify the polarization parameters of different wave systems. Using these parameters, many polarization filters are available, especially those using covariance matrices. Nonetheless, these filters only consider the three-component effect, which is essentially ineffective if there is interference between different waves. In order to overcome this unfavorable factor, a pattern recognition method is proposed. This method uses the Karhunen-Lovève (KL) transform, which is mainly composed of the following two steps: (1) Select the pattern recognition waveform, the waveform is represented by the vector subspace formed by KL’s main eigenvectors. This is a recognition step. (2) Project the entire seismic trace onto this subspace, that is, filter it. It is proposed that the above method can be applied to three-component offset VSP. The example presented shows that we can extract from all channels a signal similar to a known model waveform, such as the first downlink P wave.