利用多尺度形态学和递归图分离辨识大地电磁微弱信号

Separation and identification of magnetotelluric weak data using multi-scale morphology and recurrence plot

为了突出形态滤波中结构元素不同尺度下的相关局部特性、分层次刻画大地电磁信号本身所固有的多尺度特征,将加权多尺度形态滤波与递归图相结合对大地电磁微弱信号与噪声进行信噪分离及信噪辨识。首先,利用不同尺度的结构元素对大地电磁信号进行全方位扫描;然后,加权合成获取更精细的形态特征信息;最后,引入非线性动力学行为中的递归图对分离出的信噪特征进行确定性检验。研究结果表明:大地电磁微弱信号与噪声得到更精细分离;递归图的相空间轨迹适合定性判断大地电磁时间序列的非稳态动态变化;卡尼亚电阻率曲线更加光滑、连续,大地电磁低频段数据质量得到有效改善。

In order to highlight relevant local characteristics of different scales of structural elements for morphology filtering, and depict inherent multi-scale features of magnetotelluric data hierarchically, a new method that combined weighted multi-scale morphological filtering with recurrence plot to analyze the effect of signal-to-noise separation and identification of magnetotelluric weak data was proposed. Firstly, different scales of structural elements were used to scan magnetotelluric data comprehensively. Then, the more elaborate information of morphological characteristics was obtained by weighted synthesis. Finally, recurrence plot of non-linear dynamic behavior that inspects the characteristics of signal and noise of magnetotelluric data was adopted. The results indicate that the weak signal and noise of magnetotelluric data separate more elaborately, and recurrence plots of phase space trajectory are suitable for inspecting unsteady dynamic changes of time series of magnetotelluric qualitatively. Moreover, Cagniard resistivity curve becomes more smooth and continuous, and the quality of low frequency for magnetotelluric data is improved effectively.