基于形态和能量的大地电磁自适应高精细信噪分离方法研究

For magnetotelluric (MT) sounding field, it is an important research topic to seek modern signal processing methods to separate weak MT signal from strong interference in ore concentration area. Our preliminary research indicated that mathematical morphology filter and signal subspace enhancement, which were started from morphology and energy characteristics of strong interference, showed potential advantage in MT signal processing. It provides a new way to carry out the MT signal-to-noise separation deeply. In this project, based on preliminary research, we will focus on the theme of adaptive high precision separation. We will carry out a systematic research on effective combination of adaptive multi-scale morphological filtering, signal subspace adaptive precise decomposition, waveform mutation point precise identification and adaptive endpoint detection in MT signal-to-noise separation. We hope that the method can eliminate large scale of strong interference and baseline drift accurately when low frequency details composition of deep structural information will be maximum retained at the same time. These measures will gradually approach to the original characteristics of MT signal. Some research about the combined application of adaptive technology and signal-to-noise recognition technology in the signal-to-noise separation will be focused, and the applicability of method will be evaluated systematically. The evaluation system of signal-to-noise identification and separation for suitable ore concentration area will be established. Finally, the method of adaptive high precision signal-to-noise separation based on morphology and energy will be proposed. The research findings can provide technical support to enhance the quality of MT data in ore concentration area and improve MT sounding deep exploration ability. Moreover, it will promote the following electromagnetic inversion interpretation levels.

寻求现代信号处理方法从矿集区强干扰中分离出微弱的大地电磁（MT）信号是MT测深领域的重要研究课题。我们的前期研究表明，从强干扰的形态和能量特征出发的数学形态滤波和信号子空间增强在MT信号处理中显示出潜在的优势，这为深入开展MT信噪分离提供了思路。本项目拟在前期研究基础上，围绕自适应高精细分离这一主题，系统研究自适应多尺度形态滤波、信号子空间自适应精细分解、波形突变点精准识别和自适应端点检测在MT信噪分离中的有效结合；致力于高精度剔除大尺度强干扰和基线漂移的同时，最大限度地保留含深部构造信息的低频细节成分，逐步逼近MT信号的原始特征。关键研究自适应技术和信噪识别技术在信噪分离中的组合应用，系统评价方法的适用性，建立适合于矿集区信噪识别与信噪分离的评价体系。最终形成基于形态和能量的自适应高精细信噪分离方法，为提升矿集区MT数据品质、提高MT测深深部探测能力及后续电磁法反演解释水平提供技术支持。

强电磁干扰环境下，如何获取高质量的大地电磁数据一直是大地电磁测深领域的重要研究课题。本项目从矿集区大地电磁强干扰的形态和能量特征入手，开展了自适应多尺度形态滤波、子空间精细分解、压缩感知重构算法、局域均值分解等噪声压制方法研究。同时，为了有效保留大地电磁低频段的缓变化信息不被“过处理”，引入分形、递归和聚类等新的信号处理手段，对大地电磁信号和强干扰开展了信噪辨识研究，并结合信息论、数学形态学和非线性动力学，初步建立了大地电磁信噪辨识与信噪分离的评价准则。研究结果表明，所研方法能有效剔除大尺度强电磁干扰，微弱的大地电磁有用信号得到了更为精细的保留，低频段的数据质量得到明显改善；为后续自动高精度甄辨微弱的大地电磁有用信号和典型强干扰，以及对辨识的强干扰采取针对性的措施进一步提升低频段的大地电磁信噪分离精度奠定了基础。本项目的研究成果为在强干扰区开展大地电磁测深提供了强有力的技术支持，具有重要的理论意义和工程实用价值。