论文部分内容阅读
地震学研究中地震震相初至拾取处于基础而又关键的环节,其拾取速度和精度直接影响其在地震精确定位、震相识别、震源机制及破裂过程、地震勘探以及地震层析成像中的应用效率和精度.早期的震相初至拾取是人工的、非实时分析;随着计算机技术、数据采集与处理技术以及定量或数字地震学的发展,震相初至拾取也由早期的人工分析过渡到人机互动的半自动分析以及后来的自动实时检测.目前,可以进行震相初至自动检测的方法有很多,但没有一种单独的方法能在所有不同类型震源、传播路径、接收方式以及噪声背景下对初至进行一致的拾取,且对于信噪比低、初动不明显或后期弱震相埋在早期震相尾波中、噪声与地震信号频率相近时的地震记录,初至自动拾取的效果通常都不理想.鉴于此,有必要对目前流行的各种震相初至自动识别检测方法进行归纳总结,以期对该领域的发展有所裨益.本文就目前常用的检测方法技术(如能量法、分形维数法、频率法、偏振法、自回归模型、相关法、小波变换、人工神经网络法等)按时间域、频率域、时频域、综合方法四大类进行了回顾、分析及综述.结果表明:寻找一种综合信号和噪声多特性差异及多震相特征量的方法可能是目前初至自动检测技术的发展方向,即充分利用信号与噪声在运动学、动力学、频谱特征、偏振属性等方面的显著差异性,形成一套同时具有算法简单、检测精度高、多道处理功能、可用于实时处理特征的综合识别检测方法技术.
Seismological study in the first phase of the earthquake to pick in the basic and crucial part of the pick-up speed and accuracy of a direct impact on their precise positioning, phase identification, focal mechanism and rupture process, seismic exploration and seismic tomography Application efficiency and accuracy Early phase to phase pick-up is a manual, non-real-time analysis; with the development of computer technology, data acquisition and processing techniques and quantitative or digital seismology, Semi-automatic analysis of the transition to human-computer interaction and subsequent automatic real-time detection.At present, there are many ways to start the seismic phase automatic detection, but there is not a single method that can be used in all different types of sources, propagation paths, reception methods and In the background of noises, the first arrival is picked up consistently, and for the low signal-to-noise ratio, initial inaction, or the late weak phase buried in the early phase coda wave, the seismograms recorded when the noise is close to the seismic signal frequency, The effect is usually not ideal. In view of this, it is necessary to summarize the presently prevailing methods of automatic identification detection In order to benefit the development of this field.This paper analyzes the commonly used detection methods (such as energy method, fractal dimension method, frequency method, polarization method, autoregressive model, correlation method, wavelet transform, artificial neural network, etc. ) Are reviewed, analyzed and summarized in four categories: time domain, frequency domain, time-frequency domain and synthesis method.The results show that finding a method that combines multiple features of multi-seismic signal and noise features and multi-seismic facies features may be the first To the development of automatic detection technology, which takes full advantage of the signal and noise in the kinematics, dynamics, spectral characteristics, polarization properties of the significant differences between the formation of a set of both simple algorithm, high detection accuracy, multi-channel processing, Integrated recognition and detection techniques for real-time processing of features.